The UK Productivity “Puzzle” in an International Comparative Perspective – Fernald – 2025 – Oxford Bulletin of Economics and Statistics

1 Introduction

Slow economic growth in the UK has attracted considerable attention in recent years, with a particular focus on the role of slow productivity growth. Although productivity growth has slowed across advanced economies since the mid-2000s, there is a sense that the “productivity puzzle” has hit the UK harder than elsewhere (e.g., [1, 2]). In this paper, we analyse the UK productivity experience relative to that in the United States and major northern-European countries.¹ Northern Europe turns out to be a reasonable comparison for the UK; the experience of southern Europe is quite different.² As our analysis relies on comparable industry-level data, we are limited to the pre-pandemic period.³

Figure 1 illustrates the nature of the puzzle by showing cumulative market-economy labor-productivity growth since 1985 for the UK, U.S., and northern Europe (the “EU-5”). The figure suggests that, from 1985 until about 2000, labor productivity grew at similar rates in the three regions. After 2000, the EU-5 fell back. The UK continued to keep pace with the U.S. until 2007 but has since lost considerable ground. Although the US has pulled ahead, labor productivity growth in all regions was slow for at least a decade before the pandemic.

In this paper, we argue that the widespread slowdown after the mid-2000s was largely driven by a common slowdown in growth in total factor productivity (TFP). Overall TFP growth since 2007 has been similarly slow across regions. Indeed, given the widely discussed TFP slowdown at the U.S. “frontier,” it would have been surprising if the UK as well as northern Europe had not seen a slowdown. The residual UK TFP puzzles since the mid-2000s are modest and idiosyncratic.

So how do we explain that the UK slowdown in labor productivity growth was sharper than in U.S. or continental Europe? Growth accounting for the 1995–2007 period shows that UK TFP was converging towards the EU-5 level—which in turn, is stable and below the U.S. TFP level. The UK’s TFP convergence came to an end after 2007, thereby showing a sharper slowdown than in the EU-5 or U.S.; the importance of this TFP slowdown is also emphasised by Goodridge and Haskel [5].

Importantly, in our main accounting decomposition, we do not find that a shortfall of capital deepening contributed to the post-2007 labor-productivity slowdown. Van Reenen and Yang [6] and popular discussion argue that the UK’s low investment rate contributes to weak levels and growth of UK labor productivity. We agree. But the UK, as well as northern Europe, had relatively low investment rates since the 1990s. The shortfalls in labour productivity growth did begin after the Global Financial Crisis in the 2000s. But from 1995 to 2007, the slow pace of capital deepening was not as pressing a concern because the relatively fast pace of TFP growth at the time drove fast labor productivity growth. And while policies to stimulate investment, as argued for by Alayande and Coyle [7] and Van Reenen and Yang [6] (amongst others), would provide a boost to UK labor productivity growth, it seems implausible to expect a return to pre-2007 labor productivity growth rates without also a return to higher TFP growth rates.

The organising principle for our analysis is the logic of conditional convergence. That logic implies that countries should eventually approach a steady-state path where each country has its own level of productivity (GDP per hour or TFP) relative to the frontier. The relative steady-state level of TFP depends on conditioning or “institutional” factors such as the efficiency of financial markets (which can affect relative resource misallocation), the availability of skilled managers, or the nature of competition.⁴ If those conditioning factors don’t change, then countries should all (eventually) tend to grow at similar rates. Conversely, if countries can improve these conditioning factors, they can grow more quickly than the frontier during the transition.⁵

This conditional-convergence perspective is the received wisdom for explaining why, in the aftermath of the second world war, TFP growth in Continental Europe was initially well above U.S. or UK rates. Those rates gradually declined as Europe recovered from the destruction and dislocations of the war. As TFP levels got closer to U.S. levels, growth naturally slowed.

Conditional convergence is easier to see in levels than in growth rates. Figure 2 shows the U.S., UK, and EU-5 levels of market economy labor productivity (left panel) and TFP (right panel) since the mid-1980s.⁶ The figure illustrates several key points. First, for both labor productivity and TFP, the U.S. has remained the overall “frontier” economy throughout this period. (We note in Section 4 that the U.S. is not the TFP leader in all sectors; and idiosyncratic sectors such as mining explain much of the overall shortfall in levels.) Second, although the U.S. has been gradually pulling ahead in terms of labor productivity in the left panel (as already shown in Figure 1), TFP levels in the right panel are much more stable. For example, the EU-5 level of TFP was 90% of the U.S. level in 1985, 91% in 1995, and 91% in 2019.

Third, the UK succeeded in closing the TFP gap somewhat with the U.S. and EU-5 between 1995 and 2007, albeit unevenly. For example, UK TFP was 84% of the U.S. level in 1985 and 1995, peaked at 89% in 2008 and then fell to 86% in 2019. Finally, as the difference in scale on the vertical axis between the two panels makes clear, only part of the differences in levels of labor productivity can be traced to differences in TFP; inputs of (produced and human) capital must also contribute, as emphasised by Van Reenen and Yang [6]. Still, as we discuss in Sections 2 and 3, changes in capital intensity (capital relative to output) do not explain the productivity slowdown in the UK, U.S., or EU-5.

A key question for the UK and other countries is what happens when TFP growth at the frontier changes? Regardless of whether one focuses on conditional convergence of labor productivity (in the Solow model) or of TFP (in an endogenous growth framework), if frontier growth changes, growth everywhere should change. Though there may be a lag, the ideas and innovations that drive TFP growth diffuse across countries. Consistent with this view, the International Monetary Fund [9] uses cross-country panel data from 1970 to 2007 and finds a statistically significant link between changes in U.S. TFP growth and in TFP growth for other advanced economies. The peak effect occurs in three to 4 years. The UK experience is broadly consistent with this hypothesis that TFP growth follows frontier growth with a lag. When U.S. TFP growth picked up in the mid-1990s, UK TFP growth picked up within a few years. And when U.S. TFP growth slowed in the mid-2000s, UK TFP growth slowed within a few years.⁷

Given that the U.S. has been pulling away over time in terms of labor productivity but not TFP, there is a role for capital formation to account for some of the cross-country differences in labor productivity trends. But relatively weak UK and EU-5 investment, and the resulting relatively slow growth in capital accumulation and capital intensity, is a structural issue going back to at least the early 1990s; see also Alayande and Coyle [7] for a detailed discussion. Furthermore, following the global financial crisis (GFC) that began in 2007, we view slow growth in capital input as reflecting the endogenous response to slow growth in TFP and in demand. Indeed, capital-output ratios contributed more to growth, not less, after 2007 and a comparison of returns to investment suggest that returns are not that different from those in the U.S. As a result, stories such as Van Reenen and Yang [6] that focus primarily on capital deepening to explain the productivity slowdown are, in our view, focused on a secondary symptom rather than the primary driving factor.

Thus, the common mid-2000s TFP slowdown is central for understanding the experience of the UK, northern Europe, and elsewhere. We discuss alternative stories in Section 5. We consider the leading story to be a slowing trend that predated the Great Recession [10, 11]. The temporary mid-1990s boost to TFP growth from the production and use of information and communications technology (ICT), which has been documented in U.S. data (e.g., [10]) ended a decade later. The trend then reverted to a slower pace. That section also discusses policy implications that follow from this or alternative (demand-driven) stories.

This paper is structured as follows. Section 2 documents that the common labor-productivity slowdown was a TFP slowdown. In our growth accounting, we do not see weak capital formation as an important independent channel for explaining the post-2007 slowdown in labor-productivity growth. Section 3 focuses in more detail on the role of investment. Section 4 looks at sectoral TFP detail to try to understand the sources of TFP convergence in the 1995–2007 period. Section 5 discusses alternative hypotheses regarding the common TFP slowdown after the mid-2000s and discusses policies. Section 6 concludes, drawing some lessons for the debates.

2 Key Facts and Conceptual Framework

This section shows that weak TFP growth is the key to understanding the UK productivity slowdown after 2007, just as it is in United States and Europe. In contrast, a shortfall of capital formation after 2007 does not appear to be an important independent contributor to the productivity slowdown. Certainly, UK investment rates are low. But the challenge of weak UK investment is a longstanding problem going back to the early 1990s, not just a post-GFC problem. Section 3 discusses the role of capital formation in greater detail.

3 What Is the Role of Weak Investment?

Our view that capital deepening does not have an important independent role in explaining the UK, U.S. or European labor-productivity slowdown after 2007 may seem surprising. It is “conventional wisdom” that investment grew more slowly after 2007 than before. But the relevant input for production is capital. For productivity growth, it is growth in capital relative to output in our preferred decomposition that matters, and in the UK that ratio declined less after 2007 than it did in the fast-growth 1995–2007 period.

This section provides additional evidence that there was no apparent break in the capital formation process after 2007, apart from the endogenous response to slow TFP growth. For at least a decade prior to the Great Recession, commentators noted low levels of UK investment. Concerns about inadequate investment are somewhat alleviated by considering a broader set of intangible assets. And internal rates of return (especially when broader intangibles are included) do not look that far out of line with U.S. returns.

We start with timing: Why did the big decline in the contribution of the UK capital-output ratio to growth occur in the 1990s, not later? To provide insight, Figure 3 shows nominal national-accounting ratios of investment (gross fixed capital formation) relative to GDP since 1970. The top panel includes residential real estate; the bottom panel is only non-residential investment. These series cover the total economy, not just the market economy, because it allows us to discuss a long time period; we discuss the market economy below, and the main lessons remain.

The U.S. shows only a modest downward trend over this 50-year period. And in the 1970s and 1980s, the graph does not suggest any apparent shortfall in UK investment relative to continental Europe or the U.S. Indeed, the non-residential investment rate was often higher in the UK than elsewhere.

But after peaking in 1989 (UK) or 1990 (EU-5), both the UK and EU-5 show sharp declines in investment rates—with the UK showing a particularly sharp decline over the next decade and a half. The UK non-residential investment rate averaged 19½ percent from 1970 to 1990; it fell more than 3 percentage points, to 16¼ percent, in the 1991–1995 period. By the 2003–2007 period, the rate had fallen nearly 3 percentage points further, to only 13½ percent. Following the Global Financial Crisis, the investment rate fell a bit further for several years. But, especially for non-residential investment, the post-GFC decline is small, with the investment rate on average only about a half percentage point below its 2003–2007 period rate. By 2014, the rate had returned to its pre-GFC level. The decline is a bit more pronounced for overall investment because of the sharp GFC-induced decline in residential investment, but the general trends are similar.¹²

With sharp declines in the investment-GDP ratio in the 1990s, it is no surprise that the capital-output ratio contributed less to growth after the mid-1990s than before (in fact, it subtracts from growth after 1995). And with a relatively stable, if low, investment rate after 2007, it is no surprise that the capital-output ratio subtracts less from growth, not more, in the post-GFC period.

Of course, even if investment was already weak, how do we explain the modest decline in the investment-GDP ratio in the years right after the GFC? In this regard, it is important to differentiate cyclical declines in the investment rate from more structural ones. Investment is the most procyclical component of aggregate expenditure, in a deep recession, we expect investment to fall relative to GDP.¹³ Indeed, with output falling during the GFC, the actual capital-output ratio rose sharply (see Appendix B); that implies that the marginal product of capital was low immediately after the recession. In the years that followed, with limited TFP growth and a slow recovery in demand, firms had little need to invest in additional capacity so the investment rate recoveries gradually to its pre-GFC level.

That said, even if investment weakness does not contribute independently to the slowdown in labor productivity growth after 2007, it is still a cause for concern. In particular, the UK has had persistent, structural weakness in capital formation. This has long been recognised in analyses of UK productivity. For example, IMF [16] explicitly focuses on the low UK investment-to-GDP ratio and discusses reasons (including distortions) that might explain the low rate.¹⁴ That study, as well as Crafts and O’Mahony [17], notes that the low relative level of UK capital per hour contributed to low UK standards of living.

We are not aware of a clear consensus on the reasons for the relatively low rates of National Accounts (NA) investment that started in the 1990s. Alayande and Coyle [7] survey a range of hypotheses, including a shift away from investment-intensive manufacturing; changes in capital allowances after 1984; distinctive financial structures in the UK that discourage fixed investments; inadequate complementary human capital; inadequate complementary infrastructure and low-quality management; and persistently elevated policy uncertainty. Alayande and Coyle [7] argue that, even though it is challenging to identify the role of each specific reason, “there is little downside in addressing the issues in [their] list” (p. 22).

In this discussion, we have so far focused on NA investment. However, NA measures of gross fixed capital formation in the NA may not capture all investments, notably omitting many intangible assets (e.g., [18]; Haskel and Westlake [19]). The asset boundary of the NA has shifted over time to also include investments in such assets as software and research and development. But investment in, say, new financial products, new product design or organisational capital are still outside the NA asset boundary.

In Figure 4, we show, in the bottom three lines, the investment rate in the market economy based on NA assets for the period since 1995 (so this comparison does not speak to the early 1990s slowdown in investment). These show a broadly similar trend and cross-country pattern as in Figure 3 for non-residential investment in the total economy, with the UK falling behind the U.S. and EU-4 (investment data for Belgium are incomplete). The timing of the investment shortfall is a bit later for the market economy than for the overall economy. But of course, with faster TFP growth in the 1995–2005 period than the U.S. or the EU-5, the level of capital formation was still insufficient to prevent a fall in growth in the capital-output ratio after 1995 (Table 1).

But if we include estimates of additional intangible assets from the EU KLEMS and INTANprod database, we arrive at the top three lines. UK investment does not fall as far behind in the years after the GFC and is ahead of EU-4 investment rates almost continuously.

When extending the asset boundary, it is not just investment that changes but also output, since the additional assets have to be produced. This implies that labor productivity growth based on an extended asset boundary is also different and the EU KLEMS and INTANprod database has a set of growth accounts that makes all these adjustments. In the case of the UK, these adjustments slightly raise the rate of labor productivity growth and decrease the TFP growth rate, but the overall pattern of growth rates over time and across countries is quite similar.¹⁵ From this we conclude that taking a broader perspective on intangibles reduces the UK underinvestment argument without changing the argument that the core of the UK productivity growth problem is that TFP growth declined in tandem with the U.S. and EU-5.

As a caveat, this argument captures only the direct effect of intangibles on the accounting. The indirect effects are less clear. For example, the analysis by Corrado et al. [20] reaches a similar result for the direct growth accounting effects. However, they argue that spillovers to growth from the accumulated stock of intangibles can explain much of the mid-2000s U.S. TFP slowdown. To get this result, they assume very large and immediate positive spillovers from a broad class of intangibles—well beyond R&D—such as advertising, training, and organisational capital. We are less persuaded that spillovers to this large class of intangibles is as large as assumed by Corrado et al. [21]. For example, the argument for spillovers from organisation capital are weaker than for R&D as organisation capital tends to be more tacit, and the evidence is likewise weaker (e.g., [22]). In addition, several of the models discussed in Section 5 (e.g., [23]) suggest that intangibles may have negative spillovers by serving as a barrier to entry that slows innovation and growth.

The comparison of investment rates in Figure 4 suggests that the argument of persistent UK underinvestment is not as clearcut as the NA figures imply. But, as Van Reenen and Yang [6] argue, because the UK lags behind the US and EU-5 in terms of labor productivity levels (see Figure 2), it should actually have a higher investment rate to close the gap.

So, why is investment not higher? At a macroeconomic level, we can look at the implied return to capital, since businesses need to be incentivised to invest by the returns on that investment. We can distinguish between two broad sets of explanations by looking at these returns over time. First, it could be that the return on investment is high but there are frictions (such as high firm market power) or high risk-perception (e.g., policy uncertainty due to Brexit) that keep firms from investing. Second, it could be that returns are normal (or low), in which case current levels of capital input presumably reflect low TFP or weak demand.¹⁶ Appendix C provide a more formal exposition of this argument. Given the different root causes, the policy implications of these two broad rationales differ.

Here $$$ {K}_{i,t} $$$ is input of asset $$$ i $$$, $$$ {p}_{it}^K $$$ is the relevant investment deflator, which expresses each asset in current-cost terms, and $$$ {\delta}_i $$$ is the geometric depreciation rate of asset $$$ i $$$. We implement two versions of the IRR, one based on the set of NA assets, $$$ {IRR}^{NA} $$$, and one where the set of assets also includes additional intangible assets, $$$ {IRR}^{INT} $$$.¹⁸ Appendix C discusses how a persistent shortfall in capital relative to its optimal level will lead to an elevated IRR. For example, excessive risk premia (often discussed in the post-financial-crisis period), capital wedges such as capital taxes or other investment frictions, or monopoly profits, all can lead to non-only too-low levels of capital (i.e., “underinvestment”) but also an elevated IRR. It is difficult to measure the socially optimal levels of the user-cost of capital or the IRR, so we focus on whether the UK is out of line with the U.S.

In Figure 5 we plot our two measures of return, the left panel based only on NA assets and the right panel including additional intangible assets. The left panel suggests that returns in the UK are consistently high–above the EU-4 and US levels for (almost) all years and by a sizable 1.9–4.6 percentage points on average. If we take the U.S. levels as the benchmark, this figure is consistent with a chronic shortfall in capital. But the gap widens only modestly after 2007 and returns to roughly the same pre-GFC gap within a few years. Hence, when we focus on NA intangibles, the message is the same as in the growth accounting: Any shortfall in capital is longstanding and did not obviously get worse after the GFC.

However, after accounting for the larger investments in intangible assets and the associated income, the UK-U.S. difference shrinks considerably. There is still a gap with the U.S. level—on average, 0.8 percentage points. This persistent gap is again consistent with longstanding frictions of some sort that might lead to a chronic shortfall in capital formation; but it could also reflect measurement uncertainties (e.g., the reliability of the non-NA intangibles). In any case, the gap in returns between the UK and U.S. has shrunk in the post-2007 period—so if the UK has an investment problem that pushes up the net IRR, the U.S. does as well.

This analysis is close in spirit to Karabarbounas and Neiman [24], who define “factorless income” as income that cannot be attributed to (NA) assets earning a risk-free rate of return. They argue that rising factorless income could be due to rising intangibles (Case K); rising profits from market power (Case Π); or financial factors, such as financial frictions or changes in risk premia, that increase required returns (Case R). The link between market power and factorless income can be direct when rising market power leads to pure economic profits; but market power does not need to lead to profits and factorless income, if the rising market power is needed to offset rising returns to scale, say, from fixed costs.¹⁹ Determining whether it is market power, risk or another friction that is behind the relatively high net IRR in the UK is an interesting question, but well beyond the scope of our analysis of the UK productivity slowdown.

In sum, the story of underinvestment in the UK is more nuanced than some of the literature suggests. Although lower investment cannot account for the slowdown in labor productivity growth since 2007, in an accounting sense, higher investment would have contributed to faster labor productivity growth, closing the growing gap between the UK and U.S. (and, to a lesser extent, EU-5). However, when taking investment in a broad set of intangibles into account, UK investment levels are not particularly low in relative terms; and the net IRR is not particularly high relative to U.S. levels—and the gap in returns does not widen after 2007. This, again, points to the importance of TFP as the key underlying factor in the slowdown, not a shortfall in investment.

4 TFP Across Industry Groups

Looking at the overall market economy from a conditional convergence perspective, as we do in Figure 2, it is not at all clear that the UK has much of an overall TFP puzzle. Frontier TFP growth slowed after the mid-2000s, so it is no surprise that UK (as well as EU-5) TFP growth slowed down as well. Of course, the UK would have benefited had it been able to further close the TFP gap in levels. But it would have been a surprise if the sharp global slowdown were not associated with a UK TFP slowdown, possibly with a lag. That leaves unanswered the question of why frontier TFP growth slowed down, which is a question we return to in Section 5 below.

Relative to the U.S., the residual TFP growth “puzzles” for the UK are (i) the 0.4 percentage-point-per-year convergence that took place from 1995 to 2007 (Table 1, line 7; and the right panel of Figure 2) and (ii) the 0.25 percentage point divergence that took place since then. We look at industry (groups) to gain insight, at least in an accounting sense, into the sources of this convergence and divergence. As a first step, we divide the market economy into three groups: manufacturing; market services; and ‘other industries,’ which includes agriculture, mining, utilities, and construction. Figure 6 shows the TFP levels within each group, where the US level in 1995 is set equal to 100. On average during this period, manufacturing’s share in UK market economy value added was 20%, market services accounted for 63% and the remaining 17% consisted of agriculture, mining, utilities, and construction.

Before discussing the results of this exercise, it is important to highlight that with ONS [25], the UK National Accounts introduced ‘double deflation’ for value added at constant prices, bringing the UK into line with international statistical practice. Under this method, industry (gross) output and intermediate inputs are deflated with their deflators. This was a change from the previous single deflation approach, whereby the growth of value added at constant prices was computed using the deflator for gross output. In addition, the ONS introduced hedonic deflators, which increased growth in for telecommunication services and for clothing. These revised growth series impact the period since 1997. This change in method does not affect aggregate growth but substantially changes the distribution of growth across industries see (e.g.) Fernald et al. [26] for a more extensive discussion and comparison. As a result, some of the cross-industry patterns that researchers highlighted based on the pre-ONS [25] data, have changed considerably, see also Martin [27] and Coyle and Mei [28]. For example, average annual TFP growth in manufacturing based on the old, single-deflation method for the period 1997–2020 was only 1.1%; the new double-deflated data put the number at 2.9%. The move to double deflation should lead to better international comparability. At the same time, a degree of caution is in order as growing experience at the ONS with this new method could lead to further refinements in the future.

With these observations and caveats in mind, we highlight three broad takeaways from Figure 6. First, UK manufacturing caught up to the US level by the early 2000s; since 2007, the UK manufacturing level has surpassed the US level. Indeed, by 2019, UK manufacturing TFP was about 18% above the US level. Manufacturing TFP levels in the EU-5 has fallen behind the UK and U.S. since around 2000, though it has somewhat closed the gap with the U.S. since 2007. This seems primarily due to a slowdown in U.S. manufacturing productivity growth, a trend that has been identified elsewhere but (so far) has no clear explanation [29].

Second, the level of UK market services substantially converged to EU-5 and US levels in the 1995–2007 period. Since then, market services have been largely stable and close to (albeit about 5% below) the US and EU-5 levels. Not shown, if we combine manufacturing and market services, which together account for some 83% of the UK economy, there is strong convergence of UK TFP towards US levels before 2007. This aggregate fully converges with the EU-5 level and is only a few percentage points below the U.S. level by the end of the sample, as the higher level in manufacturing almost (but not quite) offsets the slightly lower level in market services. (Market services is, of course, more than three times as large as a share of the economy.).

But as Figure 2 showed, the overall UK level of TFP remains well below the US or EU-5. If it is not manufacturing or market services, then it must be in the remaining 17% of the market economy that we labelled “other” (agriculture, mining, utilities, and construction). Indeed, a third takeaway from Figure 6 is that the level of TFP in other industries has collapsed since the early 2000s, dropping from approximately two-thirds of the US level to only one-third.

When we look at this grouping in more detail, the UK is primarily different from the U.S. and EU-5 because of a strong decline in TFP in mining. This is likely because North Sea oil is becoming more challenging and expensive to extract. As Barnett et al. [30] note, “North Sea oil and gas extraction output has been in secular decline since around 2003.” Quantitatively, mining TFP growth has been substantially negative since the late-1990s—the same measured inputs lead to reduced output. In contrast, for the U.S., fracking was an important contributor to a pickup in mining TFP growth from 1.8% during 1995–2007 to 3.0% thereafter.

Brandt, Schreyer, and Zipperer [31] and Byrne, Fernald, and Reinsdorf [32] argue that both the U.S. and UK mining TFP numbers are, conceptually, biased downwards relative to true technological change. Consider the accessibility of an oil deposit as the “quality” of the natural resource as an input. Holding fixed that natural resource quality, suppose that the same quantity of other capital and labor leads to the same output. Then technology and TFP are both unchanged. But if the quality of the natural resource gets worse (e.g., the North Sea runs out of oil), then the same observed inputs, with unchanged technology, leads to lower output. Although technology has not changed, measured TFP falls. From this perspective, the observed decline in UK TFP in mining (oil extraction) presumably reflects the declining quality of North Sea deposits. If, despite the increased extraction costs, it is still profitable to extract and sell the oil, then there is nothing obviously problematic. In the U.S., fracking is a technological innovation that substantially lowers the cost of extraction at a given location. Because of that, new locations that were previously uneconomic are now worth drilling. In other words, fracking allowed a given quantity of observed inputs to lead to increasing amounts of oil and gas extraction, despite a shift to lower-quality deposits. Hence, measured TFP (which does not account for the shift to high cost, ‘low quality’ deposits) also understates the true technology gains. Whether this effect of deposit quality on measured mining TFP is more pronounced in the UK or U.S. is beyond the scope of this paper. The important point is that, from a macroeconomic perspective, mining appears to be a relatively self-contained and idiosyncratic issue, not a broader reflection of UK competitiveness.

As noted above, manufacturing in the UK is a relatively smaller share of the market economy than EU-5 manufacturing. This difference in economic structure can account for some of the differences in market economy TFP growth but has little effect on the magnitude of the slowdown. Counterfactual UK market economy TFP, applying EU-5’s value added shares but keeping UK industry TFP growth unchanged, would have been consistently faster. From 1995 to 2007, counterfactual growth would have averaged 1.79% rather than the observed 1.53%; and after 2007, counterfactual growth would have been 0.13% rather than the observed −0.09%.²⁰ Note that the slowdown after 2007 would have been very similar under this counterfactual, −1.66 versus the actual −1.62, again underscoring the broad-based nature of the slowdown. This confirms earlier arguments, by Mason et al. [15], that differences in economic structure are not a prime factor in understanding the UK’s productivity performance; the UK slowdown is large regardless.

As a final piece of analysis, we take an initial look at the possible effects of Brexit. This is not aimed at being dispositive; the increases in trade restrictions between the UK and EU only came into effect after the UK left the EU on January 31, 2020. Even though UK productivity growth data are available through 2021, this is likely too short a period to find a substantial effect, also given the concurrent effects of the Covid pandemic. At the same time, the Brexit referendum in 2016 may have had a chilling effect on innovative activities in industries more exposed to Brexit, anticipating more trade frictions and thus an (effectively) smaller market to innovate and produce for. To assess whether there is such a prima facie case for a Brexit effect, we split our 12 industries into those more exposed to Brexit’s effects and those less exposed based on whether an industry’s share of exports in gross output is above or below the median.

Figure 7 shows one TFP index for the aggregate of the more-exposed industries (mining, manufacturing, transport and storage, information and communication, finance and insurance, and business services) and one for less-exposed industries (agriculture, utilities, construction, wholesale and retail trade, hotels and restaurants, other services). The first thing to note is that if we view this through a difference-in-difference lens (more vs. less exposed, post vs. pre-referendum), then a parallel trends assumption is violated: The more-exposed industries were growing strongly after 1995, whereas TFP has fallen for less-exposed industries. Since the global financial crisis of 2007 and the start of the productivity slowdown, differences are less pronounced. But even in this post-2007 period, more-exposed industries have shown a larger increase in TFP growth since the referendum than less-exposed industries. This result holds also if we distinguish industries by the share of trade in value added from the OECD TiVA database rather than gross exports.²¹ This should not be seen as an argument that productivity will be unaffected by Brexit, but rather that in the period we analyse, Brexit has not been a deciding factor for understanding the slowdown in TFP growth of UK industries.

To summarise this section, we can trace the convergence of UK market economy TFP before 2007 to convergence in manufacturing and in market services. Since 2007, TFP growth slowed down markedly across industries in the UK, US and EU-5. The reason for the UK losing some ground appears to reflect primarily idiosyncratic industry issues, such as the decline in mining TFP, rather than a sign of general weak competitiveness. The results comparing TFP growth in industries more and less exposed to Brexit similarly suggests that even this very UK-specific event has not (yet) had a UK-specific impact on TFP growth across industries. Indeed, the only really clear pattern is that there has been a broad-based slowdown that can be seen in the UK, U.S. and EU-5 and across nearly all industries (also at a more detailed level than the three main groups). This result raises two questions. First, how can we understand this synchronised downturn in productivity growth across advanced economies? And second, what scope is there for UK policy to shift productivity growth to a higher trajectory?

5 Discussion and Policy Implications

Our story for the overall UK and EU slowdown in the 2000s is the common global TFP slowdown. We have, at least implicitly, taken the U.S. as the overall frontier, although it is not the overall leader everywhere. Since there is no reason to expect a frontier slowdown to lead to faster convergence by other regions, that slowdown naturally led to a slowdown everywhere.

This argument takes the high-level view that TFP growth in a country and industry depends on the diffusion of knowledge and ideas from a frontier economy—a framework that is common in the endogenous growth literature (e.g., [33]). More recent and comprehensive modelling approaches similarly trace the flow of knowledge and ideas across countries, for example, based on flows of international trade (e.g., [34, 35]). These models all provide a basis for relating TFP growth at the frontier to TFP growth in other countries; they make it clear that TFP is not manna from heaven but requires explicit investments in innovation. Still, there is no easy one-to-one mapping from these models to the substantial differences in cross-country TFP trends across detailed industry groups (Figure 6). This could be due to industry-specific factors, as with mining, or more general omitted variables, such as intangibles.

Analysts, including us, typically take the U.S. as the frontier because it is the overall market-economy-TFP leader. Of course, ideas can flow in both directions across borders and the U.S. is not always at the frontier in individual industries. But this does not change the basic point that countries do not exist in autarky. Innovations in one region of the world flow to other regions. Conversely, if a productivity slowdown in one region reflects slower growth in new innovation, then it is likely to be associated with a productivity slowdown elsewhere. There is a large literature looking at the U.S. slowdown, and our central premise is that it makes no sense to investigate why the UK (or continental Europe) slowed independently.

Fernald, Inklaar, and Ruzic [26] contrast two different reasons for the common TFP slowdown. One story emphasises the common shock of the Great Recession. The other emphasises a common trend slowdown that is largely independent of the Great Recession. The distinction matters because the policy implications differ. Do you focus on macroeconomic policies to smooth the business cycle, or even run a “high pressure economy”? Or do you focus on microeconomic distortions that impede innovation?

Fernald et al. [26] argue that the evidence inclines towards the ‘common-trend’ hypothesis. The ‘Great-Recession-shock’ view is particularly problematic for the United States. The mid-1990s U.S. productivity boom ended several years, at least, prior to the Great Recession—somewhere between 2004 and 2006 (see [10, 11]; and updated estimates from [36]). This U.S. slowdown was observed before the Great Recession, and professional forecasters were already at least partially accounting for it [11]. The standard story is that ICT had provided an exceptional boost to trend TFP growth in the mid-1990s and early 2000s. The waning trend plausibly reflected a pause, or end, in those exceptional gains.

One argument for a slowing trend is that “ideas are getting harder to find” [37]. Alternatively, several papers argue that information technology itself might endogenously lead to a slower pace of innovation and growth throughout the economy. These stories put the slowing U.S., UK, and European productivity trend in the context of other recent developments, such as declining firm dynamism, rising dispersion of firm-level productivity in many countries (e.g., [38]), and the growing importance of superstar firms [39].

For example, de Ridder [23] argues that intangibles linked to IT hardware and software are a form of fixed costs. Successful firms expand, which allows them to spread these fixed costs over more production. The initial expansion of these successful high-intangible firms, in turn, increases productivity growth even as concentration rates rise. But over time, another effect dominates: It is challenging (for new entrants and for existing low-intangible incumbents) to compete with the high-intangible incumbents, because of the need to invest in a large overhead intangibles stock. So innovative activity, firm turnover, and aggregate productivity growth slow.

Aghion et al. [40] provide a related endogenous growth argument linked to information technology. In their story, improvements in information technology initially boosts productivity by increasing managerial scope, which allows high-productivity/high-markup firms to expand. But the expansion of these high-productivity firms eventually deters innovation and undermines long-run growth. The reason is that a potential innovator would have to compete with a high-productivity incumbent. That prospect lowers the expected the returns to innovation. Though the de Ridder [23] and Aghion et al. [40] stories differ in their details, both suggest reasons for why information technology—a general purpose technology—might now be leading to reduced innovative activity at the frontiers of the global economy.²²

These micro-founded “supply-side” stories should apply equally to the U.S., UK and Europe. But with diffusion lags, timing of the slowdown need not be the same. If we take as given the three- to four-year lag estimated by the IMF [9], then the U.S. slowdown that appeared in the mid-2000s would hit the UK and Europe right around the time of the Great Recession.

Other supply-side stories emphasise regulatory burdens as the source of weak TFP growth. For example, Fernández-Villaverde and Ohanian [41] argue that regulation and a lack of competition have led to weak growth in Europe as well as the United States. That said, Fernald et al. [11] find no quantitative evidence that regulatory changes have a first-order impact on U.S. TFP growth. Philippon [42] argues that Europe now has more pro-competitive regulations than the United States; to the extent that product-market frictions explain why countries in continental Europe initially had lower TFP levels than the U.S., one might have expected that deregulation would subsequently help Europe to close the gap—that is, to have faster TFP growth than the U.S. That said, those gains might take time to appear. In any case, the policy implications of excessive regulatory burdens/inadequate antitrust protections are similar to those from other supply-side stories.

An alternative view is that, in fact, the Great Recession was a major shock that caused the TFP growth trend to slow. There are theoretical models in which deep recessions propagate through reduced innovation and (for a time) TFP and output growth (see the discussion in [26], for references). The Great Recession timing seems to fit the UK and EU-5. But as already discussed, the Great Recession story does not naturally match the pre-Great Recession timing of the U.S. slowdown. In addition, most of the evidence in the literature is that hysteresis in output levels arises from labor, not from productivity [26].

From the point of view of the UK productivity puzzle, both our preferred story (a frontier TFP slowdown) and the Great Recession alternative (which, similarly, slowed global TFP growth) imply that we should not be focusing on idiosyncratic UK stories. Rather, as we have argued throughout, the overall UK slowdown is not a surprise given the global slowdown.

In terms of policy recommendations, a first set is mainly macroeconomic. First, even if the Great Recession is not the main explanation for the mid-2000s productivity slowdown, it is presumably not helpful for innovation to run a cool economy. Moreover, there are welfare costs to excess unemployment and cyclical shortfalls in output growth, regardless of whether these shortfalls are also a drag on innovation, idea diffusion, and TFP growth. So keeping the economy close to full employment is a reasonable policy goal regardless of whether it plays an important role in the global TFP slowdown. Second, policy should aim to reduce uncertainty to the extent possible. Higher uncertainty is likely to raise the UK risk premium (and, therefore, the user cost of capital) and reduce investment.

Beyond macroeconomic stabilisation policies and policy predictability, is there also a role for industrial policy to change the structure of UK production towards faster-TFP-growth sectors? As already noted, industry structure matters somewhat for TFP growth over time, but it does not explain the TFP slowdown. And in the absence of market failures, subsidising manufacturing development, say, could easily reduce future UK welfare and might not even raise TFP growth. To a large degree, relative TFP across industries translates into relative prices. So if the UK specialises in slower-TFP-growth services, UK consumer welfare can benefit from a terms of trade improvement by importing ever-cheaper manufactured goods. Suppose subsidies to domestic manufacturing succeed in raising measured UK TFP growth. Then the UK would presumably lose (some or all of) the terms of trade gains: It might even start exporting the ever-cheaper manufacturing goods, while importing products whose prices were not falling. In the absence of market failures, the terms-of-trade losses combined with the rising tax burden would be expected to offset the gains. And of course, the marginal manufacturing firm that enters only because of the subsidies may well be less productive than the average, thereby reducing manufacturing TFP growth (even if the value-added share of manufacturing increases).

Our study does not provide any particular insight into whether, or how, these rationales apply to the UK. Still, stepping back, it is important to remember that the TFP slowdown occurred across advanced economies and across a broad range of industries. There is no reason to suspect that industrial policies can somehow reverse the global TFP slowdown; at best they may help non-frontier economies close the gap to the frontier a little further.

In our view, the most promising policy steps are not those aimed at somehow improving the industry structure. Rather, we would focus first on general (almost generic) steps to create an innovative- and business-friendly environment with a lower risk premium and, potentially, increased competition (lower markups). Such policies would lead firms to want higher capital-output ratios, and potentially stimulate investments in UK innovation. Such steps can help close the gap in productivity levels.

Still, to the extent that, as we think likely, the problem is a common global slowdown in innovation and TFP growth—independent of the Great Recession—the broader challenges cannot be solved by UK innovation policies alone. After all, the UK is by itself a modest part of global research efforts. Nevertheless, greater formal and informal investments in UK innovation would somewhat help the global innovation effort and might also aid in diffusion of global “best practices” to the UK. This could help close the gap in TFP levels. That said, as Figure 6 shows, the UK is a leader in manufacturing TFP and is only a modest laggard in market services; and much of the considerable shortfall in the “other industries” sector is idiosyncratic (as suggested by our earlier discussion of mining).

In terms of creating an innovation friendly environment, government policies can play an important role [43]. One set of policies involves creating sound institutions, such as the rule of law, efficient bankruptcy procedures, and appropriate antitrust/competitiveness policies. These policies ensure that entrepreneurs and innovators can appropriate the gains from their innovations, without allowing them to subsequently create barriers that restrict competition. A second set of policies involves promoting investments, most obviously in education, infrastructure, and research and development. Some of those investments may boost innovation and TFP—our main focus—but even apart from that, they can help close labor productivity levels gaps.

6 Conclusion and Implications for the Future

Across advanced economies, growth in labor productivity and total factor productivity slowed markedly after the mid-2000s. In this paper, we focus primarily on the UK TFP experience, viewed through the lens of conditional convergence. From this perspective, the bulk of the UK TFP slowdown reflects a slowing frontier trend. The residual puzzles are small. The UK was converging towards the U.S. TFP level in the pre-Great Recession period, driven by both manufacturing and market services; and manufacturing TFP levels are now above those in the U.S. and EU-5. In the post-2007 period, the modest UK divergence appears fairly idiosyncratic—mechanically driven, most obviously, by mining, where appropriate measurement is a challenge and there have been substantial idiosyncratic shocks. Hence, we put little weight on the modest post-2007 divergence.

Overall, there is still some scope for closing the gap with frontier economies outside manufacturing. Policymakers care about the level of productivity, and they should not be happy with keeping a roughly stable position relative to the United States. The fact that the level of UK TFP remains below the frontier highlights the theoretical possibility that growth could be relatively rapid if the UK could achieve renewed convergence dynamics. Addressing the issues discussed in [1, 2], such as misallocation across firms and regions, could promote these favourable dynamics.

Data Availability Statement

The data and code used to generated the tables and figures in this paper is available at https://doi.org/10.34894/S3AAMJ.