Across all OECD countries, economy-wide labour productivity grew by 1.2% in 2024, twice the 2023 pace. Despite elevated economic uncertainty, 29 countries saw gains in 2024. Yet, only 11 recorded labour productivity growth in 2024 above their pre-pandemic average (2010-19).
The United States outperformed most OECD countries, with labour productivity growth of 2.2% in 2024 while it was below 1% in over half of OECD countries. Italy (-1.4%), Japan (-1.3%), the United Kingdom (-0.7%) and Germany (-0.4%) recorded contractions, whereas Canada saw near‑zero growth and France experienced a small increase (+0.4%).
OECD economy-wide growth in multifactor productivity (MFP) – capturing how efficiently labour and capital inputs are combined – remained weak in 2024. Information and communication services as well as professional services experienced the strongest industry-level MFP growth.
Recent productivity figures, notably strong labour productivity growth in the United States and MFP gains in selected services industries, may provide early, tentative signals consistent with a productivity enhancing role of AI.
Early estimates based on preliminary data and nowcast estimates point to positive labour productivity growth across most OECD countries in 2025. Labour productivity growth strengthened in the European Union, accelerating from 0.2% in 2024 to 1.4% in 2025.
OECD Compendium of Productivity Indicators 2026
Since its launch in 2005, the OECD Compendium of Productivity Indicators has served as a key reference for internationally comparable insights into labour and multifactor productivity across OECD countries and, where possible, accession candidate countries. The 2026 edition of the Compendium sheds light on the heterogeneity hidden behind aggregate productivity figures by drawing on more detailed industry-level and firm data than in previous editions. It decomposes productivity trends across countries and industries along multiple dimensions to identify the drivers of economic growth, including labour, capital inputs, and multifactor productivity.
The report places recent changes in perspective through comparisons with benchmarks derived from productivity patterns over the last 30 years. It also examines trends in investment and explores productivity differences across and within enterprise size classes, as well as across regions within countries. In line with recent editions – which presented examples of extensions of the measurement framework to address challenges such as cyclicality (OECD, 2025[1]) – this year’s edition includes a chapter on environmentally adjusted multifactor productivity, an augmented version of multifactor productivity including natural capital as a production input and accounting for the negative externality of gas emissions.
Due to reporting lags in several key data sources, the most recent productivity statistics presented in most chapters refer to 2024. This introductory chapter nonetheless includes preliminary figures for 2025.
Labour productivity for the total economy, which measures how efficiently each hour of work translates into GDP, grew by 1.2% in 2024, when averaged across OECD countries using total hours worked as weights (Figure 1.1). This marks an improvement compared with 2023 (+0.6%) and is only slightly below the average rate of 1.3% observed during 2010-19, which, however, was itself a period of relatively weak labour productivity performance.
This growth in labour productivity unfolded against a challenging global economic backdrop, marked by geopolitical tensions, disruptions to shipping in the Red Sea, and a sharp rise in trade policy uncertainty (OECD, 2025[2]). Headwinds to trade can directly affect productivity by weakening export performance, eroding economies of scale and leading to capacity underutilisation, thereby reducing labour productivity. Reduced trade openness can also weigh on productivity indirectly by slowing the diffusion of technological knowledge, by dampening firms’ incentives to innovate, and by weakening the reallocation of labour and capital to the most productive firms (Shu and Steinwender, 2019[3]; Melitz and Redding, 2021[4]). At the same time, concerns over supply-chain concentration have spurred efforts to diversify sourcing (OECD, 2025[5]; OECD, 2026[6]), potentially entailing adjustment costs, duplication of production stages, and efficiency losses.
Reduced predictability in the trade policy environment has contributed to elevated uncertainty in OECD economies in recent years, weighing on firms’ investment decisions (Boer and Rieth, 2024[7]; OECD, 2025[8]). Investment – a key determinant of productivity performance – is shaped by the macroeconomic environment, notably aggregate demand and borrowing costs. As inflation moderated in 2024, monetary conditions eased, while mild fiscal tightening was underway in many OECD countries.
Against this backdrop, the OECD average investment rate, measured as economy-wide gross fixed capital formation as a share of GDP, declined modestly from 23.0% in 2023 to 22.6% in 2024 (see Chapter 3), remaining above the 2010-19 average but below levels seen in the years before the global financial crisis (2000-07). The mild downward tendency in the investment rate was mitigated by robust spending on ICT assets, possibly reflecting firms’ efforts to deploy artificial intelligence (AI).
Recent productivity figures may partly reflect emerging AI effects
While a growing body of literature provides micro-level evidence of the potential of AI to deliver productivity gains at the level of specific tasks (Cui et al., 2026[9]; Schwarcz et al., 2026[10]), recent discussions have highlighted the difficulty of detecting clear signs of AI-driven economy-wide productivity gains in official statistics (OECD, 2025[1]; International Labour Organization/International Labour Organization,, 2026[11]).
Key explanations for an apparent disconnect between micro insights and the aggregate picture include uneven AI adoption across industries and firms, as well as the need for complementary investments in training, managerial adjustments, and infrastructure to realise the potential of AI. Productivity as reflected in official statistics may stagnate or even fall during a transitionary period shaped by high adjustment costs, before gains eventually become visible in a pattern described as a J-curve (Brynjolfsson, Rock and Syverson, 2021[12]). Multifaceted measurement challenges, especially regarding investments in intangibles, may make it difficult for official statistics to fully capture the impact of AI (see Box 1.1).
While far from conclusive, and subject to substantial uncertainty, an optimistic reading of productivity developments in 2024 and preliminary estimates for 2025 (discussed further below) may, however, be interpreted as providing early, tentative signals consistent with a productivity‑enhancing role of AI. Overall OECD labour productivity growth accelerated in 2024 compared with 2023 (1.2% versus 0.6%) despite geopolitical tensions and wars, skill shortages, and elevated economic uncertainty.
Labour productivity growth can be decomposed into capital deepening – i.e. changes in capital input per hour worked – and improvements in multifactor productivity (MFP), which reflects changes in output not accounted for by changes in capital and labour inputs. Recent labour productivity growth in the United States and other OECD countries may in part be driven by capital deepening associated with AI adoption. AI‑induced improvements in economy-wide MFP depend on diffusion and complementary investments, including in skills. OECD economy-wide MFP growth continued to be slow in 2024, but information and communication services as well as professional services – which encompass activities such as legal services and management consulting – experienced the strongest MFP growth across industries in 2024 (see Chapter 4). AI adoption is particularly prevalent in these industries (Filippucci et al., 2025[13]; OECD/BCG/INSEAD, 2025[14]).
Yet, a nuanced view should also consider other factors that may have supported productivity growth in recent years, e.g. the acceleration of digitalisation during the Covid-19 pandemic (Criscuolo et al., 2023[15]; Dao, 2024[16]), unwinding of labour hoarding (Bodnár et al., 2025[17]), as well as increases in the tradability of many services (Benz, Jaax and Yotov, 2022[18]) and growth in digital trade (OECD, 2025[19]).
The rapid acceleration of AI advances and adoption since the early 2020s raises two key questions for productivity statistics: first, whether any AI‑related effects are already visible in measured productivity growth, and second, what measurement difficulties arise in capturing these effects within the current statistical framework.
Recent evidence on the link between AI and productivity
Most attempts to pin down the productivity effects of AI can be broadly subsumed into three categories. First, a rapidly growing body of literature adopts a micro perspective, often focused on individual workers, plants or firms, to explore how this new technology affects the efficiency of performing specific tasks. Contributions in this realm often find sizeable positive effects of AI on task-level productivity, e.g. regarding business problem-solving (Cruces et al., 2026[20]), mammography screenings (Gommers et al., 2026[21]), and customer support (Brynjolfsson, Li and Raymond, 2025[22]). Such micro studies also shed light on nuances and partially negative effects on productivity. For example, AI adoption may hamper skill acquisition (Shen and Tamkin, 2026[23]) and the most skilled and experienced workers might see small improvements in speed but also small reductions in quality.
A second group of studies takes an aggregate, forward‑looking perspective, using modelling approaches to predict the potential economy‑wide productivity effects of AI. Analyses in this vein typically start from existing estimates and assumptions about the effect of AI on individual tasks, incorporating them in general equilibrium models applied to different adoption scenarios (Misch et al., 2025[24]; Filippucci, Gal and Schief, 2026[25]). Acemoglu (2024[26]) predicts improvements in the average labour productivity growth rate by roughly 0.12 percentage points (p.p.) of the United States over ten years, whereas recent OECD work focussing on the G7 suggests gains in average annual labour productivity growth of 0.2-1.3 p.p. over a ten-year horizon (Filippucci et al., 2025[13]).
Thirdly, an emerging set of contributions focuses on ex‑post empirical assessments beyond individual tasks or groups of workers. As AI adoption accelerates, this perspective draws on newly available surveys and official industry-level productivity data to study observed productivity effects, in contrast to approaches that model future impacts. For example, an analysis of survey data covering 12 000 firms in 27 EU countries indicates that AI adoption increases firm-level labour productivity in the short run by 4% (Aldasoro et al., 2026[27]). Comparing measures of AI adoption to industry-level labour productivity trends in the United States, Çakır Melek and Miller (2026[28]) identify patterns suggesting that AI use is linked to within-industry improvements in labour productivity since 2022, while noting that these patterns are not yet broad‑based across industries. In a study relying on regressions using industry‑level data for 31 countries, Bick et al. (2026[29]) find that industries with higher firm‑level AI adoption experienced faster labour productivity growth in Europe and the United States over 2022-2024.
Measurement challenges
The central role of intangible capital in the investments required to adopt the new technology constitutes a key measurement challenge which may lead to an underestimate of productivity gains from a new technology such as AI during the transition phase (Brynjolfsson, Rock and Syverson, 2021[12]). Intangible investments, e.g. in training or adjustments of business processes and software, are imperfectly captured in firm-level balance sheets and national accounts. The 2025 System of National Accounts (SNA, 2025[30]) broadens the scope to include additional intangible assets and encourages provision of more detailed reporting, most notably regarding data, software, and AI. Yet, statistics according to SNA 2025 with sufficiently broad country coverage for comprehensive cross-country analyses will not be available before the early 2030s.
Further measurement challenges relate to cloud computing, which changes the boundary between investment and intermediate consumption (Ker, 2021[31]) and may lead to an underestimate of investment in AI. The difficulty of accurately capturing AI investment is further exacerbated by the lack of accurate deflators to adjust for quality changes of AI assets and insufficient information on retirement and depreciation profiles of AI assets (Bontadini et al., 2026[32]).
Beyond these hurdles concerning investment, further open questions relate to the fact that the value created by AI may in many cases not be directly observable in classic financial transactions. In order to assign a monetary value to AI output that is quasi-free for consumers or bundled with other digital services, Brynjolfsson et al. (2026[33]) conduct online choice experiments to elicit consumers’ willingness to accept compensation for giving up access to AI services. Their results point to a substantial increase in AI-driven welfare gains between July 2025 and March 2026 in the United States.
Moreover, measures of AI adoption are not harmonised across countries, limiting comparability (Bick et al., 2026[29]). Even within a given country, different surveys can yield varying estimates of adoption, for example due to differences in how AI is defined (Allen, 2026[34]).
In addition, the expansion of AI is accompanied by growing energy use and emissions, particularly through data centres and computing infrastructure. However, these environmental costs are not yet incorporated into standard productivity measures. While significant efforts have been made to account for environmental externalities in productivity analysis, existing approaches each come with their own strengths and limitations (see Chapter 6). Moreover, environmental accounting practices remain predominantly national in scope, despite the fact that AI usage in one country can give rise to emissions elsewhere, reflecting the location of server and data centre infrastructure (Coyle and Poquiz, 2025[35]).
For recent discussions of AI-related measurement challenges and potential ways of addressing them, see Coyle and Poquiz (2025[35]), Fonteneau et al. (2025[36]) and Makridis and Brynjolfsson (2026[37]).
A 2.2% increase in labour productivity in 2024 marked the second consecutive year of growth in the United States (Figure 1.1). This gain in US labour productivity in 2024 is more than double the average growth observed between 2010 and 2019. These strong gains could point to emerging effects of AI: the United States has seen AI adoption rates rise sharply, with employment-weighted firm AI adoption reaching around 78% in November 2025 (Allen, 2026[34]). The US labour productivity growth rate in 2024 (+2.2%) reached a level close to the average (+2.3%) observed for this economy during the period before the global financial crisis (2000-07).
In contrast, the European Union saw labour productivity grow by only 0.2% in 2024 – far below the average pace seen during 2010-19 (1.2%). The stronger increase in labour productivity in the United States compared with the European Union in 2024 has further widened the productivity gap across the two sides of the Atlantic (see Chapter 2). Measured in US Dollars in constant 2020 Purchasing Power Parities, labour productivity in the European Union reached only around three quarters of the US productivity level in 2024. Similar patterns are observed across other OECD economies, with Japan and Korea slightly below two thirds and Australia at just over four fifths of the US level.
GDP per hour worked, total economy, per cent
Note: For the OECD and European Union aggregates, the series start in 1995 due to data limitations for some countries regarding earlier years.
Source: OECD Productivity database.
Over half of OECD countries saw gains in labour productivity in 2024, with the rest experiencing stagnation or declines. Twenty-nine countries experienced higher labour productivity growth in 2024 compared with 2023. However, only 11 countries recorded labour productivity growth in 2024 exceeding their pre‑pandemic average (2010-19), even though this represents a relatively modest benchmark.
The labour productivity growth seen in the United States (+2.2%) in 2024 stood out among the G7 countries. Italy (-1.4%), Japan (-1.3%), Germany (-0.4%), and the United Kingdom (-0.7%) recorded contractions, whereas Canada saw near‑zero growth and France experienced a small increase (+0.4%).
Considerable heterogeneity characterised labour productivity developments across EU countries in 2024. Poland (+5.1%), Bulgaria (+4.4%) and Denmark (+3%) emerged as the top performers across the OECD. Conversely, Croatia (-1.3%) and Romania (-2.9%) recorded sizeable declines (Figure 1.2). Diversity in outcomes was also visible beyond Europe: Australia recorded a decline (-0.7%), whereas Korea (+1.8%) and Mexico (+2%) registered improvements in labour productivity.
GDP per hour worked, total economy, per cent
Source: OECD Productivity Database.
In 2024, labour productivity growth varied widely across industries in OECD countries, with mining and agriculture showing the largest cross-country dispersion. Overall productivity growth in manufacturing reflected the influence of key industries. Pharmaceutical productivity surged in Denmark in 2024, driven by GLP-1 therapies. This single industry pushed overall Danish manufacturing productivity growth well above the OECD median (see Chapter 4).
Aggregate labour productivity growth in 2024 was primarily driven by improvements within industries rather than by shifts in hours worked across industries (Figure 1.5). The within-industry effect, which captures productivity changes occurring inside industries, accounted for the majority of overall productivity growth. By contrast, the between-industry effect, reflecting changes in the distribution of total hours worked across industries, remained modest in most countries. Yet, the picture emerging from this analysis may partly reflect the chosen level of aggregation. Some reallocations of hours may go unobserved, as it can occur within industries across more detailed activities.
In percentage points
Note: The between-industry effect refers to the reallocations of hours worked between industries with different productivity levels or productivity growth rates. The within-industry effect reflects labour productivity developments that are not the result of reallocation of hours worked between industries. It is measured by the labour productivity growth in each industry weighted by the industry share in total value added (Chapter 4). For the United States, these three effects do not fully sum to the labour productivity growth due to the use of Fisher volume indices for GVA. However, the residual is typically small, especially when averaged over several years (see the methodological note). The decomposition is based on data at the A38 industry level for most countries, while data at the A21 industry level are used for Germany, France, Japan, Korea, Latvia, Poland, Portugal, Romania, Spain, and Sweden.
Source: Authors’ calculation based on OECD Productivity Database.
Preliminary data suggest labour productivity grew in most OECD countries in 2025
Early estimates based on Annual National Accounts data point to positive labour productivity growth across most OECD countries in 2025. On average, labour productivity growth strengthened in the European Union, accelerating from 0.2% in 2024 to 1.4% in 2025. Ireland, where the activities of multinational enterprises complicate the interpretation of conventional productivity measures (see Box 2.1 in Chapter 2), recorded the strongest increase, followed by Latvia and Poland (Figure 1.4, Panel A). By contrast, only a few countries – such as Italy, Luxembourg and Portugal – saw stagnant outcomes or a decline in labour productivity.
When data from Annual National Accounts is not available, estimates are derived using Quarterly National Accounts data or a nowcasting approach, as described in Dorville et al. (2025[38]). These estimates suggest a moderation in labour productivity growth at the total economy level in the United States, from 2.2% in 2024 to 1.7% in 2025 according to data from the U.S. Bureau of Economic Analysis (BEA) and the Bureau of Labor Statistics (BLS) data, close to the OECD nowcast estimate for that year. Nowcast estimates suggest that labour productivity growth in 2025 is likely to have been close to 1.3% in Japan and 1.8% in Korea (Figure 1.4, Panel B). However, preliminary estimates for 2025 are uncertain and should be interpreted with caution.
GDP per hour worked, per cent
Note: Labour productivity growth estimates for 2025 are based on OECD Annual National Accounts data. When these data are not available, estimates are derived using either OECD Quarterly National Accounts (QNA) data or nowcasting techniques as described in Dorville et al. (2025[38]) based on data up to April 2026. 95% confidence intervals are reported to reflect the uncertainty surrounding the nowcast estimates. For the United States, quarterly data are drawn from the US Bureau of Economic Analysis (BEA) and the Bureau of Labor Statistics (BLS).
Source: Authors’ calculations based on OECD Annual National Accounts, OECD Quarterly National Accounts, BLS, BEA and OECD Productivity database as well as nowcasts relying on methods described in Dorville et al. (2025[38]).
These preliminary signs of positive labour productivity growth in 2025 likely reflect two main supportive forces across many countries. First, macroeconomic policies and improved financial conditions together with gains in real labour incomes strengthened demand, cushioning the adverse effects of rising trade barriers and geopolitical tensions. Second, enthusiasm about AI sparked growth in investment and trade related to ICT equipment and software (OECD, 2025[2]). At the same time, employment and labour force participation remained stable in most countries, alongside still-tight labour markets (OECD, 2025[39]).
Early evidence suggests industry-level labour productivity developments varied across economies in 2025, likely reflecting country differences in firm dynamics, technology adoption, and business environment conditions. Given the debate about potential productivity gains from AI, the relatively strong productivity growth in information and communication services in the European Union (+3.7%) is noteworthy. Yet, the much lower growth recorded in this industry in the European Union in 2024 (+0.4%) suggests it is too early to view this as a sustained trend.
Figure 1.5 shows labour productivity growth across selected industries, highlighting the heterogeneity behind aggregate EU figures by contrasting them with Italy – which recorded one of the weakest performances – and Poland, one of the strongest performers in 2025. In Italy, growth in information and communication services appears to have been a key contributor to overall productivity performance in 2025. By contrast, labour productivity growth remained muted in manufacturing, where output declined broadly between 2023 and 2025 amid lower Chinese demand for luxury goods and reduced exports to the United States following tariff increases. Labour productivity growth was negative in professional services, an industry that in Italy is characterised by limited competitive pressures linked to high regulatory barriers to entry and expansion (OECD, 2026[40]).
A different picture emerges in the case of Poland, where sizeable gains were recorded in trade, transport and accommodation services. Amid unemployment rate below 3% and high vacancy rates, the difficulty of expanding employment may have incentivised firms in this industry to seek efficiency gains. Poland also experienced strong labour productivity growth in professional services (Figure 1.5), an industry whose contribution to total exports has been growing (OECD, 2025[41]).
Gross value added per hour worked, per cent
Note: Trade, transport, accommodation encompasses wholesale and retail trade, transportation, accommodation and food services. Professional services include professional and administrative service activities.
Source: Authors’ calculation based on the OECD Annual National Accounts database.
Aggregate productivity figures conceal differences across firms of different sizes (see Chapter 5). In 2024, small and medium-sized enterprises (SMEs; defined as firms employing 1 to 249 employees) in the business sector operated at significantly lower labour productivity levels than large firms (with 250 or more persons employed). Across OECD and accession candidate countries, unweighted average SME productivity was only 65% of that of large firms, measured by turnover per person employed. The size-productivity gap varies substantially across countries. Whereas SMEs in Ireland on average achieved only 40% of the labour productivity of large firms, Swiss SMEs’ average productivity was only 2% lower than that of large firms. Since 2013, this gap has widened in most countries (20 out of 32) and across most industries. Yet, there is considerable variation among firms of similar sizes, with around 95% of productivity variance within a given industry occurring within size classes, according to evidence from the OECD MultiProd project (see Chapter 5).
Labour productivity trends reflect the interplay of global and country-specific factors
Labour productivity outcomes ultimately reflect the joint influence of global factors and country‑specific determinants, encompassing both long‑standing structural features and cyclical developments. Key examples of well-established country-level determinants of productivity performance include skills, quality of institutions, macro-economic stability, and competition (OECD/APO, 2022[42]).
Weak investment is often considered as a key explanation for the prolonged slowdown in productivity growth observed across OECD countries in recent decades (André and Gal, 2024[43]). Germany, for example, has seen low investment in knowledge-based capital. Productivity growth in this country has also been hampered by the slow adoption of digital technologies, barriers to firm entry and exit, and a growing concentration of employment in larger and older incumbent firms in recent years (OECD, 2025[44]).
Furthermore, skill shortages constitute a major challenge for productivity growth, with recent OECD analysis indicating that addressing gaps in adult skills between OECD countries could lead to substantial improvements in average OECD labour productivity (OECD, 2024[45]). At the same time, demographic change heightens the need for life-long learning and greater labour mobility for workers of all ages (André, Gal and Schief, 2024[46]).
Differences in countries’ exposure to changes in international competition and technological shifts also contribute to variation in productivity performance. For example, the transition to electric vehicles and growing competition from Chinese producers have challenged European automotive companies (EPRS, 2024[47]). In Italy, motor vehicle production volumes fell by about 40% between November 2023 and November 2025 (OECD, 2026[40]). Adjustment costs, e.g. through retraining of workers, may weigh on current productivity performance. However, in the medium term, restructuring could support productivity gains over time through capital deepening, innovation and the exit of less productive firms.
Diverse productivity outcomes in 2024 also reflect differences in countries’ positions in the business cycle. Labour productivity tends to be procyclical, with labour hoarding – i.e. firms paying for more hours of work than would be strictly needed for current levels of output – playing a central role in explaining this pattern (Lewis and Villa, 2026[48]). Unemployment rates were at historically low levels in many OECD countries in 2024 (OECD, 2025[39]) and in several European countries, such as Austria and France, employment and hours worked held steady or even increased despite subdued economic growth (see Chapter 2). While retaining workers during periods of subdued activity can help firms respond quickly when demand recovers, it may temporarily reduce labour utilisation and contribute to weaker labour productivity growth.
Differences in business‑cycle positions are evident in the experiences of Korea and Romania, which saw diverging labour productivity outcomes in 2024. In Korea, labour productivity grew by 1.9% in 2024, coinciding with an economic rebound. Following a period of weak economic growth linked to global overcapacity in semiconductor supply, GDP growth picked up in 2024, partly supported by a recovery in global demand for chips (OECD, 2024[49]). This contrasts with the case of Romania, where labour productivity declined by 2.9% in 2024. Amid weak investment in a context of elevated policy uncertainty and subdued export performance, GDP growth fell sharply from 2.3% in 2023 to 0.9% in 2024 (OECD, 2026[50]).
MFP growth remained weak in 2023 and 2024
Multifactor productivity (MFP) growth, the primary engine of long-term productivity gains, remained weak in 2023 and 2024, continuing the slowdown seen in advanced economies since the mid-2000s (André and Gal, 2024[43]; Fernald, Inklaar and Ruzic, 2025[51]). Unweighted average MFP growth across 21 countries with available data in the period of 2023-24 is estimated at around -0.3%, compared with 0.8% in the pre-COVID period (2010-19) and about 1.5% during the period of 2000-07 (see Chapter 4). Positive MFP growth suggests improvements in efficiency, allowing more output to be generated from a given combination of labour and capital inputs.
Roughly half of countries recorded positive growth, and the remainder saw a decline. The Slovak Republic continued to display the strongest MFP growth, averaging roughly 3% in the years 2023 and 2024 (see Chapter 4). The United States recorded average MFP growth of 1.4% over these two years. The largest declines in MFP were observed in Estonia and Luxembourg, where MFP fell by around 5% and 2.5%, respectively.
The long-run slowdown in MFP growth has been linked to several potential explanations, such as low investment after the global financial crisis, demographic change, slowing economic globalisation, reduced knowledge diffusion among firms, and growing market power of intangibles-rich firms (Adler et al., 2017[52]; Akcigit and Ates, 2021[53]; Berlingieri et al., 2024[54]; De Ridder, 2024[55]). However, regarding variation over short time horizons, it is also important to bear in mind that productivity measures, including MFP, often exhibit cyclical fluctuations. This pattern is consistent with the well-documented procyclical nature of MFP (Comin et al., 2025[56]).
A decomposition of drivers of GDP growth into contributions from labour, capital and MFP for 21 countries shows that, amid sluggish MFP growth, increases in labour inputs have gained importance. During 2000-19, contributions from labour, capital, and MFP were typically more evenly shared than in the most recent period. Labour input growth has since become the dominant positive contributor in many economies, while the contribution of MFP growth was negative in 14 out of 21 countries in 2024 (see Chapter 2). The broader implication is that the post-pandemic period has not restored the earlier, more balanced growth model. In many economies, output expansion has come to depend on adding hours rather than on raising the productivity of those hours. Addressing drivers of growth beyond increases in labour inputs, including training requirements linked to population aging (André, Gal and Schief, 2024[46]), will therefore be important going forward (OECD, 2026[57]).
Differences in MFP growth across OECD countries within specific industries have become more pronounced since the 2010s. This divergence is most visible in digital-intensive industries, possibly reflecting uneven cross-country diffusion of advanced technologies (see Chapter 4). Consistent with this interpretation, the analysis of investment in ICT assets and R&D shows greater cross-country dispersion in 2024 than during the 2010-19 period (see Chapter 3).
Moreover, large and persistent regional disparities in firm-level MFP performance, observed, for example, in Italy and Spain, are evident even within the same industry and firm size classes, underscoring the role of local factors beyond firm size and industrial composition (see Chapter 5).
Extending the measurement framework to incorporate environmental pollution casts productivity developments in a new light
Conventional MFP metrics disregard the environmental pressures caused by emissions of greenhouse gases and air pollutants, and the depletion of natural capital (e.g. energy resources). To address this limitation, efforts have sought to incorporate emissions and natural capital into a growth accounting framework and derive environmentally adjusted multifactor productivity (EAMFP; see Chapter 6). The latest OECD estimates of EAMFP at the total economy level by Cárdenas Rodríguez et al. (2023[58]) rely on an augmented framework that treats emissions as undesirable byproducts of economic activity and includes natural capital alongside labour and produced capital as factor inputs. GDP can similarly be adjusted to provide a measure of output net of pollution. The efficiency gains captured by EAMFP are estimated to explain roughly half of pollution-adjusted GDP growth in 38 OECD countries over 1996-2018.
Despite significant progress in recent years, EAMFP indicators continue to face important methodological and data challenges. Increasing the timeliness of inputs will help to shed light on more recent developments in EAMFP. Further methodological work is needed to broaden the coverage of natural capital and pollution emissions, and to strengthen the estimation of shadow prices for pollution.
References
[26] Acemoglu, D. (2024), “The simple macroeconomics of AI”, Economic Policy, Vol. 40/121, pp. 13-58, https://doi.org/10.1093/epolic/eiae042.
[52] Adler, G. et al. (2017), “Gone with the headwinds: Global productivity”, IMF Staff Discussion Note, SDN/17/04, Vol. 17/04, p. 1, https://doi.org/10.5089/9781475589672.006.
[53] Akcigit, U. and S. Ates (2021), “Ten facts on declining business dynamism and lessons from endogenous growth theory”, American Economic Journal: Macroeconomics, Vol. 13/1, pp. 257-298, https://doi.org/10.1257/mac.20180449.
[27] Aldasoro, I. et al. (2026), “AI adoption, productivity and employment: Evidence from European firms”, BIS Working Papers, No. 1325, https://www.bis.org/publ/work1325.htm.
[34] Allen, J. (2026), “Monitoring AI Adoption in the U.S. Economy”, FEDS Notes, 2026-04-03, https://doi.org/10.17016/2380-7172.4032.
[43] André, C. and P. Gal (2024), “Reviving productivity growth: A review of policies”, OECD Economics Department Working Papers, No. 1822, OECD Publishing, Paris, https://doi.org/10.1787/61244acd-en.
[46] André, C., P. Gal and M. Schief (2024), “Enhancing productivity and growth in an ageing society: Key mechanisms and policy options”, OECD Economics Department Working Papers, No. 1807, OECD Publishing, Paris, https://doi.org/10.1787/605b0787-en.
[18] Benz, S., A. Jaax and Y. Yotov (2022), “Shedding light on the drivers of services tradability over two decades”, OECD Trade Policy Papers, No. 264, OECD Publishing, Paris, https://doi.org/10.1787/d5f3c149-en.
[54] Berlingieri, G. et al. (2024), “LAST BUT NOT LEAST: LAGGARD FIRMS, TECHNOLOGY DIFFUSION, AND ITS STRUCTURAL AND POLICY DETERMINANTS”, International Economic Review, Vol. 66/2, pp. 595-627, https://doi.org/10.1111/iere.12748.
[29] Bick, A. et al. (2026), “Mind the Gap: AI Adoption in Europe and the US”, CESifo Working Papers, Vol. No. 12584, https://www.ifo.de/en/cesifo/publications/2026/working-paper/mind-gap-ai-adoption-europe-and-us.
[17] Bodnár, K. et al. (2025), “Holding on: labour hoarding and firms’ expectations”, ECB Economic Bulletin 8/2025, https://www.ecb.europa.eu/press/economic-bulletin/focus/2026/html/ecb.ebbox202508_06~0e7cbefbfa.en.html.
[7] Boer, L. and M. Rieth (2024), “The Macroeconomic Consequences of Import Tariffs and Trade Policy Uncertainty”, IMF Working Papers, Vol. 2024/013, p. 1, https://doi.org/10.5089/9798400265143.001.
[32] Bontadini, F. et al. (2026), “Is Software Eating the World? Measuring the Progress and Diffusion of AI”, Mimeo, https://conference.nber.org/conf_papers/f234032.pdf.
[33] Brynjolfsson, E. et al. (2026), “What is Generative AI Worth?”, Mimeo, Stanford Digital Economy Lab, https://digitaleconomy.stanford.edu/app/uploads/2026/04/WhatsGenAIWorth.pdf.
[22] Brynjolfsson, E., D. Li and L. Raymond (2025), “Generative AI at Work”, The Quarterly Journal of Economics, Vol. 140/2, pp. 889-942, https://doi.org/10.1093/qje/qjae044.
[12] Brynjolfsson, E., D. Rock and C. Syverson (2021), “The Productivity J-Curve: How Intangibles Complement General Purpose Technologies”, American Economic Journal: Macroeconomics, Vol. 13/1, pp. 333-372, https://doi.org/10.1257/mac.20180386.
[28] Çakır Melek, N. and S. Miller (2026), “A New U.S. Productivity Chapter? What Industry Data Say About AI”, Federal Reserve Bank of Kansas City, Economic Bulletin 11 February, https://www.kansascityfed.org/research/economic-bulletin/a-new-us-productivity-chapter-what-industry-data-say-about-ai/.
[58] Cárdenas Rodríguez, M. et al. (2023), “Environmentally adjusted multifactor productivity: Accounting for renewable natural resources and ecosystem services”, OECD Green Growth Papers, No. 2023/01, OECD Publishing, Paris, https://doi.org/10.1787/9096211d-en.
[56] Comin, D. et al. (2025), “Revisiting Productivity Dynamics in Europe: A New Measure of Utilization-Adjusted TFP Growth”, Journal of the European Economic Association, https://doi.org/10.1093/jeea/jvaf003.
[35] Coyle, D. and J. Poquiz (2025), Making AI Count: The Next Measurement Frontier, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w34330.
[15] Criscuolo, C. et al. (2023), “The Role of Telework for Productivity During and Post COVID-19”, Economie et Statistique / Economics and Statistics 539, pp. 51-72, https://doi.org/10.24187/ecostat.2023.539.2097.
[20] Cruces, G. et al. (2026), “Does Generative AI Narrow Education-Based Productivity Gaps? Evidence from a Randomized Experiment”, NBER Working Papers, No. 34851, https://doi.org/10.3386/w34851.
[9] Cui, K. et al. (2026), “The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers”, Management Science, https://doi.org/10.1287/mnsc.2025.00535.
[16] Dao, M. (2024), “Post-pandemic Productivity Dynamics in the United States”, IMF Working Papers, Vol. 2024/124, p. 1, https://doi.org/10.5089/9798400279713.001.
[55] De Ridder, M. (2024), “Market Power and Innovation in the Intangible Economy”, American Economic Review, Vol. 114/1, pp. 199-251, https://doi.org/10.1257/aer.20201079.
[38] Dorville, Y. et al. (2025), “Towards more timely measures of labour productivity growth”, OECD Statistics Working Papers, No. 2025/01, OECD Publishing, Paris, https://doi.org/10.1787/436ecbb5-en.
[47] EPRS (2024), The future of European electric vehicles, European Parliamentary Research Service, https://www.europarl.europa.eu/RegData/etudes/IDAN/2024/762873/EPRS_IDA(2024)762873_EN.pdf.
[51] Fernald, J., R. Inklaar and D. Ruzic (2025), “The Productivity Slowdown in Advanced Economies: Common Shocks or Common Trends?”, Review of Income and Wealth, Vol. 71/1, https://doi.org/10.1111/roiw.12690.
[13] Filippucci, F. et al. (2025), “Macroeconomic productivity gains from Artificial Intelligence in G7 economies”, OECD Artificial Intelligence Papers, No. 41, OECD Publishing, Paris, https://doi.org/10.1787/a5319ab5-en.
[25] Filippucci, F., P. Gal and M. Schief (2026), “Aggregate Productivity Gains from Artificial Intelligence: A Sectoral Perspective”, AEA Papers and Proceedings, Vol. 116, pp. 31-35, https://doi.org/10.1257/pandp.20261035.
[36] Fonteneau, F. et al. (2025), “Advancing the measurement of investments in artificial intelligence”, OECD Artificial Intelligence Papers, No. 47, OECD Publishing, Paris, https://doi.org/10.1787/13e0da2f-en.
[21] Gommers, J. et al. (2026), “Interval cancer, sensitivity, and specificity comparing AI-supported mammography screening with standard double reading without AI in the MASAI study: a randomised, controlled, non-inferiority, single-blinded, population-based, screening-accuracy trial”, The Lancet, Vol. 407/10527, pp. 505-514, https://doi.org/10.1016/s0140-6736(25)02464-x.
[11] International Labour Organization/International Labour Organization, (2026), The aggregation paradox of AI, ILO, Geneva, https://doi.org/10.54394/00034342.
[31] Ker, D. (2021), “Measuring cloud services use by businesses”, OECD Digital Economy Papers, No. 304, OECD Publishing, Paris, https://doi.org/10.1787/71a0eb69-en.
[48] Lewis, V. and S. Villa (2026), “Labor productivity, effort and the Euro Area business cycle”, European Economic Review, Vol. 183, p. 105223, https://doi.org/10.1016/j.euroecorev.2025.105223.
[37] Makridis, C. and E. Brynjolfsson (2026), “Counting AI: A blueprint to integrate AI investment and use data into US national statistics”, Brookings, https://www.brookings.edu/articles/counting-ai-a-blueprint-to-integrate-ai-investment-and-use-data-into-us-national-statistics/.
[4] Melitz, M. and S. Redding (2021), Trade and Innovation, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w28945.
[24] Misch, F. et al. (2025), “AI and Productivity in Europe”, IMF Working Papers, Vol. 2025/067, p. 1, https://doi.org/10.5089/9798229006057.001.
[57] OECD (2026), Foundations for Growth and Competitiveness 2026, OECD Publishing, Paris, https://doi.org/10.1787/40a7532f-en.
[6] OECD (2026), Global value chain repositioning: Insights from the 2023-24 TiVA nowcasting exercise, OECD Publishing, Paris, https://doi.org/10.1787/8c97068d-en.
[40] OECD (2026), OECD Economic Surveys: Italy 2026, OECD Publishing, Paris, https://doi.org/10.1787/539538b2-en.
[50] OECD (2026), OECD Economic Surveys: Romania 2026, OECD Publishing, Paris, https://doi.org/10.1787/4844067e-en.
[19] OECD (2025), Deriving experimental estimates of digital trade, OECD Publishing, Paris, https://doi.org/10.1787/c7dbbc14-en.
[1] OECD (2025), OECD Compendium of Productivity Indicators 2025, OECD Publishing, Paris, https://doi.org/10.1787/b024d9e1-en.
[8] OECD (2025), OECD Economic Outlook, Volume 2025 Issue 1: Tackling Uncertainty, Reviving Growth, OECD Publishing, Paris, https://doi.org/10.1787/83363382-en.
[2] OECD (2025), OECD Economic Outlook, Volume 2025 Issue 2: Resilient Growth but with Increasing Fragilities, OECD Publishing, Paris, https://doi.org/10.1787/9f653ca1-en.
[44] OECD (2025), OECD Economic Surveys: Germany 2025, OECD Publishing, Paris, https://doi.org/10.1787/39d62aed-en.
[41] OECD (2025), OECD Economic Surveys: Poland 2025, OECD Publishing, Paris, https://doi.org/10.1787/483d3bb9-en.
[39] OECD (2025), OECD Employment Outlook 2025: Can We Get Through the Demographic Crunch?, OECD Publishing, Paris, https://doi.org/10.1787/194a947b-en.
[5] OECD (2025), OECD Supply Chain Resilience Review: Navigating Risks, OECD Publishing, Paris, https://doi.org/10.1787/94e3a8ea-en.
[45] OECD (2024), Adult skills and productivity: New evidence from PIAAC 2023, OECD Publishing, Paris, https://doi.org/10.1787/8bc2c556-en.
[49] OECD (2024), OECD Economic Surveys: Korea 2024, OECD Publishing, Paris, https://doi.org/10.1787/c243e16a-en.
[42] OECD/APO (2022), Identifying the Main Drivers of Productivity Growth: A Literature Review, OECD Publishing, Paris, https://doi.org/10.1787/00435b80-en.
[14] OECD/BCG/INSEAD (2025), The Adoption of Artificial Intelligence in Firms: New Evidence for Policymaking, OECD Publishing, Paris, https://doi.org/10.1787/f9ef33c3-en.
[10] Schwarcz, D. et al. (2026), “AI-Powered Lawyering: AI Reasoning Models, Retrieval Augmented Generation, and the Future of Legal Practice”, Journal of Law & Empirical Analysis, Vol. 3/1, pp. 220-250, https://doi.org/10.1177/2755323x261427048.
[23] Shen, J. and A. Tamkin (2026), “How AI impacts skill formation”, arXiv arXiv:2601.20245, https://doi.org/10.48550/arXiv.2601.20245.
[3] Shu, P. and C. Steinwender (2019), “The Impact of Trade Liberalization on Firm Productivity and Innovation”, Innovation Policy and the Economy, Vol. 19, pp. 39-68, https://doi.org/10.1086/699932.
[30] SNA (2025), System of National Accounts 2025, United Nations, https://doi.org/10.18356/9789211073225.