## Box 1: The Tradable and Nontradable Sectors of the U.S. Economy

## Source details

**Canonical URL:** [Box 1: The Tradable and Nontradable Sectors of the U.S. Economy](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025156-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025156-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025156-source-pdf.pdf.json)

---

### Definitions and key distinctions
- Some sectors (e.g., finance and manufacturing) are described as almost entirely tradable while other sectors (e.g., hospitality, health care, and education) are described as almost entirely nontradable.
- Most sectors contain both tradable and nontradable subsectors (example: retail has tradable e-commerce vs. less tradable traditional big box stores).
- Tradability can shift over time due to changes in policies and/or technological improvements (examples cited: telemedicine, e-commerce), but such shifts tend to be slow-moving, often constrained by regulation.
- The tradable part of the U.S. economy has external sources of demand and supply, making demand less constrained because firms can sell to the 95 percent of the global population that resides outside U.S. borders.

### Implications for demand, supply, and macro dynamics
- In nontradable sectors, domestic demand and supply must coincide; nontradable consumption and investment growth are constrained by the domestic supply side.
- Tradable sectors can draw on external demand and supply, reducing domestic demand constraints.
- Structural and macroeconomic factors shifted much of the world, including the U.S., into a supply-constrained pattern of growth with accompanying inflationary pressures.
- Factors cited as constraining supply-side responsiveness and contributing to inflationary cost pressures include:
  - The end of the previous trend of rising trade-to-output ratios (post-GFC leveling of trade relative to output).
  - Aging demographics.
  - Major changes in labor market behavior.
  - Diversification in global supply networks after shocks (pandemics, wars, climate, geopolitical tensions).
- Some technologies (e.g., those enabling remote work in tradable services) have provided resiliency and growth in tradable services.
- The post-pandemic period (2021-23) exhibited persistently high inflation for the first time in several decades, attributed in part to pandemic-induced supply-demand imbalances and large fiscal programs that prevented a balance-sheet recession and supported demand.
- The primary gating factor for U.S. growth is identified as the supply side; increased migration has been beneficial to the U.S. economy in recent years.
- The anticipated AI-fueled productivity boom may not materialize or may be far more muted than current expectations.

### Empirical and literature context
- The paper builds on structural classifications and trade/productivity literatures, including Fisher (1939); De Gregorio and Wolf (1994); Amador and Soares (2017); Blinder and Krueger (2013); Jensen and Kletzer (2005); Gervais and Jensen (2019); Autor et al. (2013); Autor and Hanson (2021); Ang (2004); Azenui and Rada (2021).
- The authors apply a geographic concentration measure of tradability (Jensen and Kletzer (2005)) with adjustments and trade-statistics checks; characterized as a better proxy for domestic tradability than international tradability.

### Methodology for classifying tradability
- Alternatives considered and tradeoffs:
  - Sector-specific share of exporting firms (ignored non-exporting firms facing global competition).
  - Allocating whole sectors as tradable/nontradable (coarse).
  - Occupation task-based offshorability measures (subjective biases).
- Adopted approach:
  - Geographic concentration measure: highly geographically concentrated industries are more tradable; geographically dispersed industries are less tradable.
  - Adjusted using trade statistics and applied tradability shares to BEA industry-specific employment (full-time equivalent employees) and real value added (2017 chained dollars).
  - Real value added defined as sales minus purchased inputs; data are in 2017 chained dollars.
- Robustness checks:
  - News-based measures from Dow Jones’ Factiva database (1995-2022) tracking mentions of exports, imports, or cross-border activity showed little variation in news chatter around industry tradability for largely nontradable industries.
  - Retail appears among the most tradable within largely nontradable sectors by both geographic and news-based indices.
- Note: The geographic concentration approach is less dynamic (static) and better reflects domestic tradability; news-based measures probe dynamism.

### Key empirical list: Tradability Shares across Industries (as reported)
- Agriculture 100
- Mining 100
- Utilities 19
- Construction 0
- Manufacturing 100
- Wholesale 0
- Retail 15
- Transport & Warehousing 0
- Information 66
- Finance & Insurance 68
- Real Estate 0
- Legal 0
- Computer Systems Design 100
- Other Professional Services (incl. Consulting) 100
- Management of Companies 0
- Administrative and Support Services 19
- Waste Management 0
- Education 1
- Health Care & Social Assistance 2
- Arts and Entertainment 0
- Accommodation & Food 0
- Other Services 0
- Government 0

### Data and measurement notes
- Employment measure: full-time equivalent employees from BEA industry-specific data.
- Value added: real value added in 2017 chained dollars.
- Tradability shares applied to both employment and real value added series.

*Source: Box 1: The Tradable and Nontradable Sectors of the U.S. Economy.*

---

### 2.2 Labor Productivity and its Decomposition

### Measurement and drivers of labor productivity
- Labor productivity measured as value added per worker or full-time equivalent employed.
- Influencing factors: human capital and skills, digital capital intensity or digital footprint in a sector, tangible capital intensity, and the state of technology.
- Capital-intensive sectors (e.g., mining and certain utilities) have high measured labor productivity because they are capital intensive.

### Divisia index decomposition methodology
- Aggregate labor productivity E_t expressed as:
  - E_t = Σ_{i=1}^n P_{i,t} X_{i,t} / (P_t L_t)
  - Equivalent expression: E_t = Σ_{i=1}^n ε_{i,t} λ_{i,t} p_{i,t}
    - ε_{i,t} = sectoral labor productivity in activity i (X_{i,t}/L_{i,t})
    - λ_{i,t} = employment share of sector i (L_{i,t}/L_t)
    - p_{i,t} = sectoral terms of trade of sector i (P_{i,t}/P_t)
- Decomposition (continuous/discrete):
  - ln(E_T/E_0) = ∫_0^T Σ θ_{i,t} dln(p_{i,t}) + ∫_0^T Σ θ_{i,t} dln(ε_{i,t}) + ∫_0^T Σ θ_{i,t} dln(λ_{i,t})
  - Discrete components (weighted by sector nominal VA shares):
    - Direct-productivity effect = exp[ Σ (θ_{i,0}+θ_{i,T})/2 * ln(ε_{i,t}/ε_{i,t−1}) ]
    - Reallocation effect = exp[ Σ (θ_{i,0}+θ_{i,T})/2 * ln(λ_{i,t}/λ_{i,t−1}) ]
    - Terms-of-trade effect = exp[ Σ (θ_{i,0}+θ_{i,T})/2 * ln(p_{i,t}/p_{i,t−1}) ]
- Properties:
  - Divisia index is additive and symmetric and does not produce a residual.

### Interpretation of decomposition components
- Direct effect typically dominates aggregate labor productivity performance.
- Reallocation effect:
  - Positive when workers move from lower-than-average productivity activities to higher-than-average ones.
  - Negative when sectors lose workers relative to the rest of the economy; can be positive for aggregate productivity if those sectors were lower-than-average productivity.
- Terms-of-trade effect:
  - Positive when high-productivity sectors’ prices grow faster than the general price level, increasing their VA weight.
  - Market structure component expected to be relatively small and close to zero.

### Key results — Overall employment and value added structure (1998–2023)
- Nontradable dominance in 2023:
  - Nontradable sector accounted for 77 percent of employment (112 million full time jobs) in 2023.
  - Nontradable sector accounted for 68 percent of real value added ($15 trillion) in 2023.
- 1998 benchmarks:
  - In 1998, 72 percent of employment was nontradable.
  - In 1998, 74 percent of real value added was nontradable.
- Job changes 1998–2023:
  - Total incremental increase of 28 million jobs from 1998 to 2023.
  - 95 percent or 27 million of the incremental jobs came from the nontradable sector.
  - The tradable sector saw a small increase in jobs since 1998 of 1 million.
- Real value added change 1998–2023:
  - The change of $9 trillion in real value added between 1998 and 2023 was more evenly distributed between the tradable and nontradable sectors, with the nontradable sector accounting for 60 percent of the increase over the period.

### Trends in tradable vs. nontradable employment and value added
- Tradable employment:
  - 1998 to 2007: fell on an annualized basis by 0.5 percent per year.
  - 2010 to 2019: rose by 1.7 percent per year.
  - Since 2019: growth moderated but remained positive and outpaced growth in nontradable employment.
- Nontradable employment:
  - Grew steadily in subperiods by over 1 percent per year on average.
- Tradable real value added growth:
  - 1998 to 2007: annualized growth 3.9 percent.
  - 2010 to 2019: 2.9 percent.
  - More recently: 3.2 percent.
- Observations:
  - Tradable value-added increased faster than tradable employment and faster than nontradable value-added, producing persistent tradable productivity increases.
  - Improvements in employment growth rates contrasted with declines in real value added growth rates in recent decades, challenging productivity, especially in nontradable sectors.
  - All series show declines around the GFC and during the pandemic.

### Nontradable sector structure and trends
- 2023 composition:
  - Government is the largest nontradable employer followed by health care & social assistance.
  - Real estate, rental, and leasing is the largest nontradable value-added sector.
  - Government ranks second in value-added contribution; government value-add estimates are imperfectly measured (cost proxies).
  - For top nontradable sectors, employment shares outweigh value-added shares, except real estate, rental, and leasing which has an outsized value-added contribution relative to employment.
  - Within government activity, state & local governments account for a majority of both employment and value-added.
- Trends:
  - Healthcare, hospitality, and government employment drove longer-term nontradable employment growth.
  - Government’s contribution to nontradable employment growth fell dramatically in the decade leading up to the pandemic compared to 1998 to 2007.
  - Government real value added stagnated since 2010; real estate displaced government as the largest nontradable sector in value added.
  - Health care & social assistance real value added has steadily grown broadly in line with employment growth trends.

### Tradable sector structure and trends
- Tradable split (2023):
  - Goods-producing sectors (manufacturing plus agriculture and mining) accounted for 42 percent of tradable employment and 39 percent of tradable value added in 2023.
  - Tradable services account for 58 percent of tradable employment and 61 percent of real value added in 2023.
  - Large tradable services include professional, scientific, and technical services, and finance and insurance.
  - Information services and mining have real value added shares more than double their shares of tradable employment.
  - Within manufacturing, food, beverages, and tobacco and metal products are sizable employment sectors; chemical products, computer and electronics, and food, beverages, and tobacco account for sizable value-added sources.
- Recent trends:
  - Recent tradable employment growth came from manufacturing, computer systems & design, and professional services like finance and insurance.
  - Since 2010, both manufacturing value added and employment have trended upwards.
  - The multi-decade decline in manufacturing employment stopped and reversed in the decade leading up to the pandemic.
  - Post-pandemic: first recovery in recent history from a recession with a complete recovery to pre-crisis levels in manufacturing employment.
  - Strong tradable value-added growth also came from industries like information and computer systems & design.

### Box 2 — Pandemic impacts on tradable and nontradable
- Employment shocks in 2020:
  - Nontradable employment dropped by 7 percent in 2020.
  - Tradable employment fell by 4 percent in 2020.
- Recovery and 2019 vs. 2023 comparisons:
  - 2023 levels vs. 2019: nontradable employment was 3 percent above 2019 levels.
  - 2023 levels vs. 2019: tradable employment was 5 percent above 2019 levels.
- Sectoral pandemic effects:
  - Within nontradable sectors, healthcare and hospitality employment experienced large negative hits, especially hospitality.
  - Nontradable retail employment had already begun to decline prior to the pandemic and fell further during the pandemic, likely reflecting a rise in e-commerce.
- Value added:
  - Pandemic hit to nontradable value added larger than tradable: -3 percent vs. -2 percent from 2019 to 2020.
  - Recovery by 2023:
    - 2023 tradable value added up 14 percent vs. 2019.
    - 2023 nontradable value added up 7 percent vs. 2019.
- Manufacturing during pandemic:
  - Manufacturing value added recovered faster than employment and was above pre-pandemic levels by 2021.
  - By 2022, manufacturing employment had recovered close to its 2019 levels.
  - Manufacturing employment’s weak initial recovery driven by soft recoveries in fabricated metals, machinery, and computers & electronic products—likely affected by supply chain and labor supply pressures that have since dissipated.
- Aggregate implication:
  - Tradable industries helped buoy the U.S. economy more during the pandemic than during the GFC (tradable employment fell 10 percent from 2007 to 2010 while nontradable employment fell only 4 percent), reflecting the resiliency of the knowledge economy and its ability to function remotely.

### Tradability, teleworkability, and resiliency
- Teleworkability measure:
  - Constructed from the U.S. BLS 2021 Business Response Survey to the Coronavirus Pandemic: teleworkability = 100 minus the percentage of jobs that “rarely or never” involve teleworking.
  - Survey collected between July and September 2021; mapping choices applied for certain industries.
- Empirical correlations:
  - Resiliency (change in employment and value added between 2019 and 2023) is positively and strongly correlated with teleworkability.
  - Strongest relationship between tradable value-added growth and teleworkability; similar significant relationship between tradable employment growth and teleworkability.
  - For nontradable outcomes:
    - Relationship between teleworkability and nontradable employment growth is weak or insignificant.
    - Positive and significant relationship between teleworkability and nontradable value added.
- Interpretation:
  - Teleworkable industries experienced less unemployment during the pandemic and were more resilient because:
    - They employ higher-skill workers with more secure employment links.
    - Teleworkable tradable industries could rely on external, foreign markets to support demand.
  - Teleworkability is highest in the knowledge economy; knowledge economy is larger in the tradable sector but also present in nontradable activities (management, accounting, service delivery), helping explain some nontradable value-added resiliency.
- Examples:
  - Resilient and teleworkable tradable industries: computer systems design and related professional services.
  - Resilient and teleworkable nontradable industries: portions of the information sector, management of companies and enterprises, portions of finance and insurance.
  - Less resilient and not teleworkable nontradable industry example: arts, entertainment, and recreation.

*Source: wpiea2025156-source-pdf - 2.2  Labor Productivity and its Decomposition*

---

### 3.2 Labor Productivity Trends

### Sectoral labor productivity levels and trends (1998-2023)
- Tradable labor productivity grew at just under 3 percent (average annual growth) over 1998-2023.
- Nontradable labor productivity grew at 0.7 percent (average annual growth) over 1998-2023.
- The non-tradable sector accounts for close to 80 percent of employment.
- In 1998 the level of productivity in tradable and nontradable sectors was similar, with the nontradable sector having a slight lead; by 2021 tradable labor productivity was nearly double that of nontradable.
- The Covid pandemic caused a fall in productivity growth within both tradable and nontradable sectors in the period 2021-2022; by 2023 productivity trends had recovered to prepandemic levels.
- During the GFC:
  - Tradable employment declined by less than 200,000 full-time equivalent jobs and returned to pre-recession levels by 2013.
  - Real value added dropped in 2009 but quickly recovered above pre-recession levels.
  - Labor productivity growth had caught up to the long-term historical trend by the end of 2009.
- Nontradable labor productivity showed a steady upward slope since 1998 with less pronounced recession effects; for real estate both value added and employment fell during the recession, leaving labor productivity almost unaffected.
- Pandemic-specific effects:
  - Capital per worker rose as employment fell, increasing labor productivity in the short run.
  - Labor quality increases accounted for about two-thirds of labor productivity growth in Q2 2020, and 25 percent of that labor quality increase was due to the change in the distribution of workers across sectors (massive job losses in leisure, hospitality, tourism, and other low wage industries).

### Drivers from Divisia index decomposition (main effects and periods)
- Decomposition periods: 1998-2023, 1998-2007, 2010-2019, and 2019-2023.
- Components: Direct (within-sector) effect; Reallocation effect; Terms-of-trade effect.
- Key findings:
  - Direct effect dominates aggregate productivity growth and is mostly driven by the tradable sector.
  - Aggregate reallocation effect is negative for the first three time periods, indicating movement of labor from high-productivity to low-productivity sectors.
  - Tradable sector shows a negative reallocation effect (manufacturing shedding labor); nontradable sector shows a positive reallocation effect (gaining workers).
  - Terms-of-trade component for the tradable sector contributed negatively to aggregate labor productivity growth for most years (prices of manufacturing, information, and computer systems rising more slowly relative to the overall price level in the tradable sector).
  - Nontradable terms-of-trade patterns less clear; in 2023 nontradable terms-of-trade effect was positive, reflecting greater nominal weight of nontradable sectors in recovery.
- Temporal dynamics:
  - Late 1990s–early 2000s: productivity growth surge.
  - Pre-GFC: productivity growth fell leading up to GFC; spiked during GFC; dropped to nearly zero post-GFC.
  - 2010–2014: almost zero aggregate labor productivity growth; negative growth rate between 2012 and 2013.
  - 2016–2020: strong positive direct labor productivity growth with complementarities and spillovers.
  - Pandemic and recovery: positive reallocation effects as labor shifted to relatively higher-productivity sectors.

### Structural contributors to slowdown post-GFC
- Reduced investment due to risk aversion and tighter lending conditions.
- Lower marginal improvements in ICT technologies; aging workforce.
- Potential measurement error capturing digital goods and services.
- Structural change toward nontradable services with below-average labor productivity, explaining negative aggregate reallocation effect.
- Induced technical change hypothesis may be limited by labor abundance in some low-productivity sectors (e.g., healthcare and education) and by capital investment constraints and reduced business dynamism post-GFC.

### Sectoral patterns and notable contributions
- High value-added sectors (especially tradable services) show higher productivity growth; many high-employment, lower value-added parts of the economy—mostly nontradable—show low productivity growth.
- Information, wholesale, and retail maintained positive productivity growth even during 2010-2019.
- Manufacturing, finance, and insurance are major drivers of trends in the tradable sector; information and computer systems consistently contributed positively to within-sector labor productivity growth.
- Recent negative terms-of-trade contribution in tradable sector implies nominal weights increased for tradable sectors with relatively lower labor productivity.
- As the U.S. recovered from the pandemic, nontradable sectors became relatively more important in nominal terms.

### Labor productivity and wages (convergence patterns)
- Ratio of wages to labor productivity indicates greater wage convergence with productivity in the nontradable sector.
- In 2023 the nontradable wage-to-productivity ratio was 11 percentage points higher than for the tradable sector.
- Industry highlights:
  - Computer systems & design (tradable): ratio of wages to productivity peaked at 193 percent in 2000; fell to 49 percent in 2023.
  - Manufacturing: wage convergence declined from 71 percent in 1998 to 38 percent in 2023.
  - Nontradable high convergence sectors include education, construction, healthcare, and administrative services.
  - Real estate (nontradable): workers make only 5 percent of their estimated labor productivity in 2023.

*Source: wpiea2025156-source-pdf — 3.2 Labor Productivity Trends*

---

### 3.5 Digital Technology and Productivity

### Recent advances and economic context
- Growing digital footprint in the knowledge economy contributes to resiliency via communication, collaboration, e-commerce, and financial services platforms.
- Recent breakthroughs include AI, speech and handwriting recognition, image and object recognition, and reading comprehension and language understanding (including large language models (LLMs) and related generative AI).
- Some LLMs have improved to the point of outperforming average human performance on tests including the SATs, GRE, and legal Bar exams.
- These advances expand the range of human activities machines can now perform or outperform.

### Features of LLMs, adoption patterns, and development trajectory
- LLMs are highly accessible; low barriers to use and prompt creation (example cited: ChatGPT had 100 million users after two months).
- LLMs cover many domains and switch domains in response to prompts more effortlessly than earlier AI technologies.
- Performance improves as model sizes increase.
- Provider roadmaps include licensing LLMs and offering APIs for specialized applications.
- Largest impacts likely in knowledge-intensive sectors.

### Risks, limitations, and complementary human roles
- LLMs are prediction machines and sometimes produce bad predictions.
- Training relies on huge computing power (mostly available to large cloud companies) and large-scale internet data ingestion.
- Vulnerabilities: bias, prejudice, conspiracy theories, negative content, and hallucinations.
- In fact-sensitive applications (e.g., legal briefs), humans, search engines, and fact checking must remain part of deployment.

### Potential productivity impacts, sectoral reach, and distributional considerations
- Potential for a very large productivity surge across many sectors, especially knowledge-intensive ones.
- With a tradable vs. nontradable overlay, the productivity surge could be broad-based and cross the tradable/nontradable divide.
- Potential leveling-up effect could be fairly generalized.
- Manufacturing may benefit via robotics and embedding Gen AI in robots to increase productivity in blue-collar occupations by enabling communication and machine-to-human collaboration.
- In many nontradable sectors requiring in-person presence, current AI-based digital technology may be less applicable than in higher-productivity nontradable sectors with knowledge overlays.
- Possibility of adverse distributional effects and that the productivity surge may be less broad-based and more subdued than some anticipate.

### Box 3 — The Potential for Productivity Surges Associated with Digital Assistants (customer service example)
- Study: AI trained on many audio recordings of customer service interactions plus performance measures; deployed as an AI assistant to one set of agents while a comparable set did not have access.
- Main findings:
  - Overall productivity increase about 14 percent achieved in a very short time.
  - Performance increment present across experience levels; very large at the lower end of the experience spectrum—AI substantially reduced the performance differential across experience.
- Interpretation: AI may substitute for or speed up the learning process, producing leveling-up effects in agent performance.
- Implication: Productivity and other performance measures in many sectors may materialize and potentially reverse recent productivity trends in parts of the economy.

### Link to broader conclusions and outlook
- Nontradable service sectors continue to dominate employment in the U.S., accounting for over 70 percent of jobs, with most absolute employment gains from these domestically-focused sectors.
- Tradable sector has higher value-added growth than nontradables and, combined with modest tradable employment growth, yields consistently higher labor productivity growth in tradables, widening the productivity gap with nontradables.
- Overall productivity growth trend has been distinctly negative since the decade following the GFC through the onset of the pandemic; both high- and low-productivity sectors saw sharp drops in productivity growth.
- Part of aggregate productivity drop due to composition effects (expansion of labor share in lower-productivity sectors), but most attributable to direct productivity declines within sectors.
- Contributing factors include shortfalls in public and private investment and demand-constrained growth patterns after the GFC, with household consumption a very large fraction of aggregate demand (on the order of 68%).
- Structural changes: manufacturing employment decline stopped, leveled out and then increased modestly in the decade after the GFC; tradable services employment growth significant.
- Geoeconomic tensions and shift away from hyperglobalization imply higher costs, distortions, inefficiencies, and potential rewiring of globalization toward blocs of trusted partners.
- Near-term: possible slight acceleration in U.S. manufacturing employment growth as recent policies exert influence; long-term manufacturing output and employment effects may be negative due to lower potential growth globally.
- Forward-looking potential: powerful digital technologies (including biomedical science and energy innovations) could arrest declining productivity trends and possibly produce, by the end of the decade, a surge in U.S. productivity—contingent on effective spread and adoption by large employment sectors in the nontradable economy.
- Data and uncertainty: postpandemic structural-change data limited to a couple of years; ultimate outcomes depend on whether secular headwinds to growth and inflation prevail or diminish, and on diffusion of productivity-enhancing technologies.
- Suggested future research: deeper empirical causal analysis on the role of policy in driving sectoral shifts and how to reverse the productivity gap between tradable and nontradable sectors; cross-country and regional comparative analyses.

*Source: 3.5 Digital Technology and Productivity, from the provided IMF Working Paper content.*

### Box 1: The Tradable and Nontradable Sectors of the U.S. Economy

### Box 1: The Tradable and Nontradable Sectors of the U.S. Economy

### Definitions and key distinctions
- Some sectors (e.g., finance and manufacturing) are described as almost entirely tradable while other sectors (e.g., hospitality, health care, and education) are described as almost entirely nontradable.
- Most sectors contain both tradable and nontradable subsectors (example: retail has tradable e-commerce vs. less tradable traditional big box stores).
- Tradability can shift over time due to changes in policies and/or technological improvements (examples cited: telemedicine, e-commerce), but such shifts tend to be slow-moving, often constrained by regulation.
- The tradable part of the U.S. economy has external sources of demand and supply, making demand less constrained because firms can sell to the 95 percent of the global population that resides outside U.S. borders.

### Implications for demand, supply, and macro dynamics
- In nontradable sectors, domestic demand and supply must coincide: nontradable consumption and investment growth are constrained by the domestic supply side; when domestic demand is weak, domestic nontradable output must fall in tandem because international trade is not feasible for these goods/services.
- Tradable sectors can draw on external demand and supply, reducing domestic demand constraints.
- Over recent decades, structural and macroeconomic factors shifted much of the world, including the U.S., into a supply-constrained pattern of growth with accompanying inflationary pressures.
- Factors cited that constrained supply-side responsiveness and contributed to inflationary cost pressures include:
  - The end of the previous trend of rising trade-to-output ratios (post-GFC leveling of trade relative to output).
  - Aging demographics.
  - Major changes in labor market behavior.
  - Diversification in global supply networks after shocks (pandemics, wars, climate, geopolitical tensions).
- Some technologies (e.g., those enabling remote work in tradable services) have provided resiliency and growth in tradable services.
- The post-pandemic period (2021-23) exhibited persistently high inflation for the first time in several decades, attributed in part to pandemic-induced supply-demand imbalances and large fiscal programs that prevented a balance-sheet recession and supported demand.
- The primary gating factor for U.S. growth is identified as the supply side; increased migration has been beneficial to the U.S. economy in recent years.
- The anticipated AI-fueled productivity boom may not materialize or may be far more muted than current expectations.

### Empirical and literature context
- The paper connects to literature on structural classifications of economic activities and their implications for productivity and trade, building on:
  - Allan Fisher’s primary/secondary/tertiary classification (Fisher, 1939).
  - Export-data and occupation task-based approaches (De Gregorio and Wolf (1994); Amador and Soares (2017); Blinder and Krueger (2013)).
  - Geographic concentration approaches (Jensen and Kletzer (2005); Gervais and Jensen (2019)) which the authors adopt and adapt.
  - The ‘China shock’ literature (Autor et al., 2013; Autor and Hanson, 2021).
  - Productivity analysis frameworks (Ang (2004); Azenui and Rada (2021)).
- The authors apply a geographic concentration measure of tradability (Jensen and Kletzer (2005)) with adjustments and trade-statistics checks; the measure is characterized as a better proxy for domestic tradability than international tradability.

### Methodology for classifying tradability
- Considered alternatives and tradeoffs:
  - Using sector-specific share of exporting firms (ignores non-exporting firms facing global competition).
  - Allocating whole sectors as tradable/nontradable (coarse; ignores rising services tradability).
  - Occupation task-based offshorability measures (subjective biases).
- Adopted approach:
  - Geographic concentration measure: highly geographically concentrated industries are more tradable; geographically dispersed industries are less tradable.
  - Adjustments made by checking relevant trade statistics.
  - Applied tradability shares to BEA industry-specific data for employment (full-time equivalent employees) and real value added (2017 chained dollars).
  - Real value added defined as sales minus purchased inputs; data are in 2017 chained dollars to enable comparisons across time net of price changes.
- Robustness checks:
  - Constructed news-based measures of tradability for traditionally nontradable industries using Dow Jones’ Factiva database (1995-2022) tracking mentions of exports, imports, or cross-border activity using Boolean algorithms.
  - Found little variation in news chatter around industry tradability over time for largely nontradable industries, lending support to the static geographic approach as a first-order approximation.
  - Within largely nontradable sectors, retail appears among the most tradable by both geographic and news-based indices.
- Note: The geographic concentration approach is less dynamic (static) and better reflects domestic tradability; the authors consider news-based measures to probe dynamism.

### Key empirical list: Tradability Shares across Industries (as reported)
- SectorTradability Share
- Agriculture100
- Mining100
- Utilities19
- Construction0
- Manufacturing100
- Wholesale0
- Retail15
- Transport & Warehousing0
- Information66
- Finance & Insurance68
- Real Estate0
- Legal0
- Computer Systems Design100
- Other Professional Services (incl.  Consulting)100
- Management of Companies0
- Administrative and Support Services19
- Waste Management0
- Education1
- Health Care & Social Assistance2
- Arts and Entertainment0
- Accommodation & Food0
- Other Services0
- Government0

### Data and measurement notes
- Employment measure: full-time equivalent employees from BEA industry-specific data.
- Value added: real value added in 2017 chained dollars, used to avoid price-level distortions from inflation, including during the post-pandemic surge in inflation.
- The paper applies the tradability shares to both employment and real value added series.

*Source: Box 1: The Tradable and Nontradable Sectors of the U.S. Economy.*

### 2.2  Labor Productivity and its Decomposition

### 2.2  Labor Productivity and its Decomposition

### Measurement and drivers of labor productivity
- Labor productivity is measured as value added per worker or full-time equivalent employed.
- Influencing factors include human capital and skills, the digital capital intensity or digital footprint in a sector, tangible capital intensity, and the state of technology in the sector.
- Sectors that are capital intensive (e.g., mining and certain utilities) have high measured labor productivity because they are capital intensive.

### Divisia index decomposition methodology
- Aggregate labor productivity (E_t) is defined as the sum of weighted averages of sectoral labor productivities:
  - E_t = Σ_{i=1}^n P_{i,t} X_{i,t} / (P_t L_t)
  - Multiplying by L_{i,t}/L_t yields: E_t = Σ_{i=1}^n ε_{i,t} λ_{i,t} p_{i,t}
    - ε_{i,t} = sectoral labor productivity in activity i (X_{i,t}/L_{i,t})
    - λ_{i,t} = employment share of sector i (L_{i,t}/L_t)
    - p_{i,t} = sectoral terms of trade of sector i (P_{i,t}/P_t)
- Differentiating and integrating under continuity leads to a Divisia decomposition of aggregate labor productivity growth:
  - ln(E_T/E_0) = ∫_0^T Σ θ_{i,t} dln(p_{i,t}) + ∫_0^T Σ θ_{i,t} dln(ε_{i,t}) + ∫_0^T Σ θ_{i,t} dln(λ_{i,t})
  - Discrete expressions for the three components (weighted by sector nominal VA shares) are:
    - Direct-productivity effect = exp[ Σ (θ_{i,0}+θ_{i,T})/2 * ln(ε_{i,t}/ε_{i,t−1}) ]
    - Reallocation effect = exp[ Σ (θ_{i,0}+θ_{i,T})/2 * ln(λ_{i,t}/λ_{i,t−1}) ]
    - Terms-of-trade effect = exp[ Σ (θ_{i,0}+θ_{i,T})/2 * ln(p_{i,t}/p_{i,t−1}) ]
- The direct productivity effect captures within-sector output per worker gains (technology and capital). The reallocation effect captures worker shifts across sectors with different productivity levels. The terms-of-trade (market structure) effect captures changes in relative prices and sectoral importance.
- The Divisia index is additive and symmetric and does not produce a residual.

### Interpretation of decomposition components
- The direct effect tends to dominate aggregate labor productivity performance because reallocation and terms-of-trade are weighted averages whose average is constrained near zero.
- Reallocation effect:
  - A positive reallocation effect indicates movement of workers from lower-than-average productivity activities to higher-than-average productivity activities.
  - A negative reallocation effect at the sectoral level indicates the sector lost workers relative to the rest of the economy; this can be positive for aggregate productivity if the sector shedding labor was lower-than-average productivity.
- Terms-of-trade effect:
  - Reflects changes in sectoral nominal value-added shares driven by relative price changes.
  - A positive terms-of-trade effect arises when high-productivity sectors’ prices grow faster than the general price level, increasing their VA weight.
  - A negative terms-of-trade contribution indicates prices of outputs in high productivity sectors rise more slowly than the overall price level while lower-than-average productivity sectors’ prices rise relatively faster.
  - The market structure component for the macroeconomy is expected to be relatively small and close to zero.

### Key results — Overall employment and value added structure (1998–2023)
- Nontradable dominance in 2023:
  - Nontradable sector accounted for 77 percent of employment (112 million full time jobs) in 2023.
  - Nontradable sector accounted for 68 percent of real value added ($15 trillion) in 2023.
- 1998 benchmarks:
  - In 1998, 72 percent of employment was nontradable.
  - In 1998, 74 percent of real value added was nontradable.
- Job changes 1998–2023:
  - Total incremental increase of 28 million jobs from 1998 to 2023.
  - 95 percent or 27 million of the incremental jobs came from the nontradable sector.
  - The tradable sector saw a small increase in jobs since 1998 of 1 million.
- Real value added change 1998–2023:
  - The change of $9 trillion in real value added between 1998 and 2023 was more evenly distributed between the tradable and nontradable sectors, with the nontradable sector accounting for 60 percent of the increase over the period.

### Trends in tradable vs. nontradable employment and value added
- Tradable employment trends:
  - From 1998 to 2007, tradable employment fell on an annualized basis by 0.5 percent per year.
  - From 2010 to 2019, tradable employment rose by 1.7 percent per year.
  - Since 2019, tradable employment growth moderated but remained positive and outpaced growth in nontradable employment.
- Nontradable employment trends:
  - Nontradable employment grew steadily in both subperiods by over 1 percent per year on average.
- Tradable real value-added growth:
  - Annualized tradable real value added grew by 3.9 percent from 1998 to 2007.
  - Grew by 2.9 percent from 2010 to 2019.
  - Grew by 3.2 percent more recently.
- Observations:
  - Tradable value-added has increased at a much faster pace than tradable employment and faster than nontradable value-added, leading to persistent increases in tradable productivity.
  - Improvements in employment growth rates juxtaposed against declines in real value added growth rates across recent decades create a challenge for productivity, especially for the nontradable sector.
  - All series show declines around the GFC and its immediate aftermath, as well as during the pandemic.

### Nontradable sector structure and trends
- 2023 nontradable composition:
  - Government is the largest nontradable employer followed by health care & social assistance.
  - Real estate, rental, and leasing is the largest nontradable value-added sector.
  - Government ranks second in value-added contribution, though government value-add estimates are imperfectly measured (cost proxies).
  - For top nontradable sectors, employment shares outweigh value-added shares, except real estate, rental, and leasing which has an outsized value-added contribution relative to employment.
  - Within government activity, state & local governments account for a majority of both employment and value-added.
- Trends over time:
  - Healthcare, hospitality, and government employment drove longer-term nontradable employment growth.
  - Government’s contribution to nontradable employment growth fell dramatically in the decade leading up to the pandemic compared to 1998 to 2007.
  - Government real value added stagnated since 2010; real estate displaced government as the largest nontradable sector in value added.
  - Health care & social assistance real value added has steadily grown broadly in line with employment growth trends in the decade leading up to the pandemic and since then.

### Tradable sector structure and trends
- Tradable sector split (2023):
  - Goods-producing sectors (manufacturing plus agriculture and mining) accounted for 42 percent of tradable employment and 39 percent of tradable value added in 2023.
  - Tradable services account for 58 percent of tradable employment and 61 percent of real value added in 2023.
  - Large tradable services include professional, scientific, and technical services, and finance and insurance.
  - Information services and mining have real value added shares more than double their shares of tradable employment.
  - Within manufacturing, food, beverages, and tobacco and metal products are sizable employment sectors; chemical products, computer and electronics, and food, beverages, and tobacco account for sizable value-added sources.
- Recent trends:
  - Tradable employment growth more recently came from manufacturing, computer systems & design, and professional services like finance and insurance.
  - Since 2010, both manufacturing value added and employment have trended upwards.
  - The multi-decade decline in manufacturing employment stopped and reversed in the decade leading up to the pandemic.
  - The post-pandemic period is the first recovery in recent history from a recession with a complete recovery to pre-crisis levels in manufacturing employment.
  - Strong tradable value-added growth also came from industries like information and computer systems & design.

### Box 2 — Pandemic impacts on tradable and nontradable
- Employment shocks in 2020:
  - Nontradable employment dropped by 7 percent in 2020.
  - Tradable employment fell by 4 percent in 2020.
- Recovery and 2019 vs. 2023 comparisons:
  - 2023 levels vs. 2019: nontradable employment was 3 percent above 2019 levels.
  - 2023 levels vs. 2019: tradable employment was 5 percent above 2019 levels.
- Sectoral pandemic effects:
  - Within nontradable sectors, healthcare and hospitality employment experienced large negative hits, especially hospitality.
  - Nontradable retail employment had already begun to decline prior to the pandemic and fell further during the pandemic, likely reflecting a rise in e-commerce.
- Value added:
  - Pandemic hit to nontradable value added was larger than the hit to tradable value added: -3 percent vs. -2 percent from 2019 to 2020.
  - Both nontradable and tradable value added have bounced back:
    - 2023 tradable value added up 14 percent vs. 2019.
    - 2023 nontradable value added up 7 percent vs. 2019.
- Manufacturing during pandemic:
  - Manufacturing value added recovered faster than employment and was above pre-pandemic levels by 2021.
  - By 2022, manufacturing employment had recovered close to its 2019 levels.
  - Manufacturing employment’s weak initial recovery was driven by soft recoveries in fabricated metals, machinery, and computers & electronic products—likely affected by supply chain and labor supply pressures that have since dissipated.
- Aggregate implication:
  - Tradable industries helped buoy the U.S. economy more during the pandemic than during the GFC (tradable employment fell 10 percent from 2007 to 2010 while nontradable employment fell only 4 percent), reflecting the resiliency of the knowledge economy and its ability to function remotely.

### Tradability, teleworkability, and resiliency
- Teleworkability measure:
  - Constructed from the U.S. BLS 2021 Business Response Survey to the Coronavirus Pandemic: teleworkability = 100 minus the percentage of jobs that “rarely or never” involve teleworking.
  - Survey collected between July and September 2021; some industries not explicitly covered and mapping choices were applied for certain industries.
- Empirical correlations:
  - Resiliency (change in employment and value added between 2019 and 2023) is positively and strongly correlated with teleworkability.
  - The relationship is strongest between tradable value-added growth and teleworkability, and there is a similarly significant relationship between tradable employment growth and teleworkability.
  - For nontradable outcomes:
    - The relationship between teleworkability and nontradable employment growth is weak or insignificant.
    - There is a positive and significant relationship between teleworkability and nontradable value added.
- Interpretation:
  - Teleworkable industries experienced less unemployment during the pandemic and were more resilient because:
    - Teleworkable industries employ higher-skill workers with more secure employment links.
    - Teleworkable tradable industries could rely on external, foreign markets to support demand.
  - Teleworkability is highest in the knowledge economy (where the main input is information); the knowledge economy is larger in the tradable sector but also has a footprint in nontradable activities (management, accounting, service delivery), which helps explain some value-added resiliency in nontradable sectors.
- Examples:
  - Resilient and teleworkable tradable industries: computer systems design and related professional services.
  - Resilient and teleworkable nontradable industries: portions of the information sector, management of companies and enterprises, portions of finance and insurance.
  - Less resilient and not teleworkable nontradable industry example: arts, entertainment, and recreation.

*Source: wpiea2025156-source-pdf - 2.2  Labor Productivity and its Decomposition*

### 3.2  Labor Productivity Trends

### 3.2 Labor Productivity Trends

### Sectoral labor productivity levels and trends (1998-2023)
- Tradable labor productivity grew at just under 3 percent (average annual growth) over 1998-2023.
- Nontradable labor productivity grew at 0.7 percent (average annual growth) over 1998-2023.
- The non-tradable sector accounts for close to 80 percent of employment.
- In 1998 the level of productivity in tradable and nontradable sectors was similar, with the nontradable sector having a slight lead; by 2021 tradable labor productivity was nearly double that of nontradable.
- The Covid pandemic caused a fall in productivity growth within both tradable and nontradable sectors in the period 2021-2022; by 2023 productivity trends had recovered to prepandemic levels.
- During the GFC: tradable employment declined by less than 200,000 full-time equivalent jobs and returned to pre-recession levels by 2013; real value added dropped in 2009 but quickly recovered above pre-recession levels; labor productivity growth had caught up to the long-term historical trend by the end of 2009.
- Nontradable labor productivity showed a steady upward slope since 1998 with less pronounced recession effects; for real estate both value added and employment fell during the recession, leaving labor productivity almost unaffected.
- Pandemic-specific effects: capital per worker rose as employment fell, increasing labor productivity in the short run; labor quality increases accounted for about two-thirds of labor productivity growth in Q2 2020, and 25 percent of that labor quality increase was due to the change in the distribution of workers across sectors (massive job losses in leisure, hospitality, tourism, and other low wage industries).

### Drivers from Divisia index decomposition (main effects and periods)
- Decomposition periods reported: 1998-2023, 1998-2007, 2010-2019, and 2019-2023 (GFC years omitted from some analyses; the recent pandemic is included).
- The decomposition splits aggregate labor productivity growth into:
  - Direct (within-sector) effect: typically from technological improvements and capital investment within sectors.
  - Reallocation effect: productivity growth across sectors due to labor transfer between sectors.
  - Terms-of-trade effect: changes in relative prices across sectors.
- Key decomposition findings:
  - Contributions from the direct effect dominate aggregate productivity growth and are mostly driven by the tradable sector.
  - The aggregate reallocation effect is negative for the first three time periods (1998-2023, 1998-2007, 2010-2019), indicating movement of labor from high-productivity to low-productivity sectors.
  - Tradable sector shows a negative reallocation effect (manufacturing shedding labor); nontradable sector shows a positive reallocation effect (gaining workers).
  - Shorter periods have smaller growth compared to longer periods; bars for 2019-2023 are smaller than for 1998-2023 or 1998-2007.
  - The terms-of-trade component for the tradable sector contributed negatively to aggregate labor productivity growth for most years (prices of manufacturing, information, and computer systems rising more slowly relative to the overall price level in the tradable sector).
  - Nontradable terms-of-trade patterns are less clear, but in 2023 the nontradable terms-of-trade effect was positive, reflecting greater nominal weight of nontradable sectors in recovery.
- Temporal dynamics and notable episodes:
  - Late 1990s–early 2000s: productivity growth surge (exceptions: internet bubble burst and 9/11).
  - Pre-GFC: productivity growth fell leading up to GFC; spiked during GFC; dropped to nearly zero post-GFC.
  - 2010–2014: almost zero aggregate labor productivity growth; negative growth rate between 2012 and 2013.
  - 2016–2020: strong learning-by-doing effects and technological innovation (significant positive direct labor productivity growth) with complementarities and spillovers across sectors and positive terms-of-trade effects.
  - Pandemic and recovery: positive reallocation effects as labor shifted to relatively higher-productivity sectors; losses in hospitality, accommodation and food services, arts and entertainment; gains in finance and insurance, scientific and technical services, government, healthcare and social assistance, and transportation.
- Structural contributors to slowdown post-GFC:
  - Reduced investment due to risk aversion and tighter lending conditions; lower marginal improvements in ICT technologies; aging workforce; potential measurement error capturing digital goods and services.
  - Structural change toward nontradable services with below-average labor productivity, explaining negative aggregate reallocation effect.
  - Induced technical change hypothesis may be limited by labor abundance in some low-productivity sectors (e.g., healthcare and education) and by capital investment constraints and reduced business dynamism post-GFC.

### Sectoral patterns and notable sectoral contributions
- High value-added sectors (especially tradable services) show higher productivity growth; many high-employment, lower value-added parts of the economy—mostly nontradable—show low productivity growth.
- Information, wholesale, and retail maintained positive productivity growth even during 2010-2019, likely driven by the rise in online retail.
- Manufacturing, finance, and insurance are major drivers of trends in the tradable sector; information and computer systems consistently contributed positively to within-sector labor productivity growth.
- Empirical findings referenced:
  - Autor et al. (2013): increased Chinese import competition could account for up to one-quarter of the sharp decline in the U.S. manufacturing workforce between 1990 and 2007.
  - Herrendorf et al. (2015): fast-growing trade in services likely affects nature and speed of labor reallocation.
  - Mendieta-Muñoz et al. (2022): falling employment shares in high-productivity services such as finance, information and technology and wholesale trade; rising employment in low-productivity services like health, education, and entertainment.
- Recent negative terms-of-trade contribution in tradable sector implies nominal weights have increased for tradable sectors with relatively lower labor productivity.
- As the U.S. recovered from the pandemic, nontradable sectors became relatively more important in nominal terms.

### Labor productivity and wages (convergence patterns)
- Ratio of wages to labor productivity (weighted averages) indicates greater wage convergence with productivity in the nontradable sector.
- In 2023 the nontradable wage-to-productivity ratio was 11 percentage points higher than for the tradable sector (a reversal from the late 1990s).
- Industry breakout highlights:
  - Computer systems & design (tradable): ratio of wages to productivity peaked at 193 percent in 2000; fell to 49 percent in 2023.
  - Manufacturing: wage convergence declined from 71 percent in 1998 to 38 percent in 2023 (stalled since the end of the GFC).
  - Nontradable high convergence sectors include education, construction, healthcare, and administrative services.
  - Real estate (nontradable): workers make only 5 percent of their estimated labor productivity in 2023.

*Source: wpiea2025156-source-pdf — 3.2 Labor Productivity Trends*

### 3.5  Digital Technology and Productivity

### 3.5  Digital Technology and Productivity

### Recent advances and economic context
- Growing digital footprint in the knowledge economy contributes to the resiliency of the U.S. economy through communication, collaboration, e-commerce, and financial services platforms.
- Recent breakthroughs include AI, speech and handwriting recognition, image and object recognition, and reading comprehension and language understanding (including large language models (LLMs) and related generative AI).
- Some LLMs have improved to the point of outperforming average human performance on tests including the SATs, GRE, and legal Bar exams.
- These advances replace prior limits based on codification with sophisticated, multi-layered pattern recognition capabilities, expanding the range of human activities machines can now perform or outperform.

### Features of LLMs, adoption patterns, and development trajectory
- LLMs are highly accessible; low barriers to use and prompt creation (example: ChatGPT had 100 million users after two months).
- LLMs cover a wide number of domains and switch domains in response to questions, requests, or prompts more effortlessly than earlier AI technologies.
- Performance of LLMs continues to improve in part because model sizes are increasing dramatically.
- The technology and product roadmap for some providers includes licensing LLMs and offering APIs so firms can build specialized applications on the LLM platform.
- Because these models operate in the knowledge part of the economy, the largest impacts are likely to occur in knowledge-intensive sectors.

### Risks, limitations, and complementary human roles
- LLMs are prediction machines; they are not perfect and sometimes produce bad predictions.
- Training relies on huge amounts of computing power (available mostly to large cloud computing companies) and on large-scale internet data ingestion.
- Vulnerabilities include bias, prejudice, conspiracy theories, other negative content, and production of hallucinations (making things up).
- In applications where assertions masquerading as facts are problematic (for example, legal briefs), humans, search engines, and fact checking need to remain part of the deployment process.

### Potential productivity impacts, sectoral reach, and distributional considerations
- There is meaningful potential for a very large productivity surge across many sectors, especially knowledge-intensive ones.
- Using a sectoral lens with a tradable vs. nontradable overlay, the productivity surge could be broad-based and cross the tradable vs. nontradable divide.
- Potential positive distributional feature: a leveling-up effect that could be fairly generalized.
- Manufacturing may benefit via robotics; breakthroughs in robotics and embedding Gen AI in robots could increase productivity in blue-collar occupations by enabling communication and machine-to-human collaboration, compensating for spatial intelligence and navigation limitations of legacy robots.
- However, in many nontradable sectors where work is blue collar and requires in-person presence, current AI-based digital technology may be less applicable than in higher-productivity nontradable sectors with large knowledge-economy overlays.
- Therefore, there is a possibility of adverse distributional effects and that the productivity surge may be less broad-based and more subdued than some anticipate.

### Box 3 — The Potential for Productivity Surges Associated with Digital Assistants (customer service example)
- Study setup: AI trained on a large collection of audio recordings of customer service interactions plus performance measures (problem solved, duration, customer satisfaction), then deployed as an AI assistant to one set of agents while a comparable set did not have access to the AI.
- Main findings:
  - Overall productivity increase was about 14 percent and was achieved in a very short amount of time.
  - Performance increment was present across experience levels but was very large at the lower end of the experience spectrum—AI substantially reduced the performance differential across experience.
- Interpretation: AI may substitute for or speed up the learning process, producing leveling-up effects in agent performance.
- Implication: Productivity and other performance measures in a wide range of sectors may materialize; if strong enough, they may reverse recent trends in productivity in parts of the economy.

### Link to broader conclusions and outlook
- Nontradable service sectors continue to dominate employment in the U.S., accounting for over 70 percent of jobs, with most absolute employment gains from these domestically-focused sectors.
- Tradable sector has higher value-added growth than nontradables and, combined with modest tradable employment growth, yields consistently higher labor productivity growth in tradables, widening the productivity gap with nontradables.
- Overall trend in productivity growth has been distinctly negative since the decade following the GFC through the onset of the pandemic; both high- and low-productivity sectors saw sharp drops in productivity growth, contributing to lower aggregate productivity growth.
- Part of the aggregate productivity drop is due to composition effects (expansion of labor share in lower-productivity sectors), but most is attributable to direct productivity declines within sectors.
- Contributing factors to past declines include shortfalls in both public and private investment and demand-constrained growth patterns after the GFC, with household consumption being a very large fraction of aggregate demand (on the order of 68%).
- Structural changes: the long-term trend of declining manufacturing employment stopped, leveled out and then increased modestly in the decade after the GFC; tradable services employment growth is significant and, with positive manufacturing employment growth, the tradable sector has again begun contributing to overall employment growth.
- Geoeconomic tensions and a shift away from hyperglobalization (flows of goods and services, capital, technology, and people now subject to growing restrictions and industrial policies) imply higher costs, distortions, and inefficiencies, and a potential rewiring of globalization toward blocs of trusted partners.
- Near-term effects: possible slight acceleration in U.S. manufacturing employment growth as recent policies exert influence; long-term effects on manufacturing output and employment may be negative due to lower potential growth in the U.S. and globally.
- Forward-looking potential: powerful digital technologies (including biomedical science and energy innovations) could arrest declining productivity trends and possibly produce, by the end of the decade, a surge in U.S. productivity—contingent on effective spread and adoption by large employment sectors in the nontradable economy.
- Data and uncertainty: available postpandemic structural-change data are limited to a couple of years; ultimate outcomes are unknown and depend on whether secular headwinds to growth and inflation prevail or diminish, and on the diffusion of productivity-enhancing technologies.
- Suggested future research: deeper empirical causal analysis on the role of policy in driving sectoral shifts and how to reverse the productivity gap between tradable and nontradable sectors; cross-country and regional comparative analyses.

*Source: 3.5 Digital Technology and Productivity, from the provided IMF Working Paper content.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025156-source-pdf.pdf_