## wpiea2021166-print-pdf

## Source details

**Canonical URL:** [wpiea2021166-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021166-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021166-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021166-print-pdf.pdf.json)

---

### I. Introduction — framing and key properties of AI
- Advances in artificial intelligence (AI) and related automation technologies raise risks of job losses and increasing inequality, with developing countries and emerging market economies especially vulnerable because their comparative advantage relies on abundant labor and natural resources.
- Key properties of AI highlighted:
  - Likely to be labor-saving and resource-saving, devaluing sources of comparative advantage for many developing countries and deteriorating their terms of trade.
  - As an information technology, AI tends to give rise to natural monopolies and “superstar” firms located in a few countries that serve global markets.
  - Market-chosen rate and direction of technological progress are generally suboptimal, creating scope to steer innovation toward socially desirable directions.
- Policy framing:
  - Domestic policies and international cooperation, including rewriting global rules governing the information economy, are required to mitigate adverse effects.
  - Pillars of response include taxation, redistribution, targeted expenditure (education, infrastructure), steering technology adoption toward labor-using sectors (agriculture, services), and supra-national reforms (global tax regime, competition policy, intellectual property, information/data policy).

### II. Downside risks of technological progress — empirical and theoretical findings
- Empirical observations:
  - Episodes exist in which wages lagged productivity growth; median wages may not have kept up even where average wages did.
  - Widening income inequality in most advanced economies since the early 1980s; in the US and other high-income countries most benefits of growth have accrued to those at the top.
- Theoretical taxonomy (Hicksian and related):
  - Capital-biased technical change: decreases the relative share of labor.
  - Skill-biased technical change: decreases the share of unskilled labor.
  - Labor-saving technical change: can lead to an outright reduction in wages.
- Quasi-rents and redistribution:
  - Innovations generate quasi-rents accruing to winners (e.g., skilled workers, firm owners); governments can capture quasi-rents through taxation and redistribution, potentially with limited distortionary effects (e.g., taxing owners of fixed factors such as land).
- Cross-border implications:
  - When benefits accrue in one country while costs fall on another, Pareto improvement would require cross-border compensation; current arrangements are voluntary and limited.
  - Some countries may gain while others lose, reflected in terms-of-trade improvements and deteriorations.

### III. Labor-saving technological progress and task-level effects
- Labor-market patterns and job polarization:
  - Computerization and automation since the 1970s–1990s substituted routine tasks and increased demand for non-routine skills; these changes may explain nearly 2/3 of the relative demand shift toward college-educated labor over that period.
  - Employment in nonroutine jobs has continued to grow; employment in routine jobs has stagnated or declined since around 1990, contributing to hollowing out of the middle.
  - COVID-19 (2020) accelerated the trend temporarily, producing a large decline in employment in routine manual jobs but only a modest dip in nonroutine cognitive jobs.
- Task-level effects and IA vs AI:
  - Automation often replaces particular tasks rather than whole jobs; intelligence-assisting devices and algorithms (IA) can be complementary to labor by improving or taking over certain tasks.
  - Both AI and IA imply extensive restructuring and reallocation of worker time toward non-automated tasks.
- Distributional consequences:
  - Labor-saving progress can make the world richer overall but disproportionately harm developing countries with a comparative advantage in cheap labor, potentially causing deterioration in terms of trade and large export-income losses.
  - Resource-saving progress can reduce demand for depletable natural resources and harm resource-exporting countries (e.g., oil-exporters), while benefiting some resource-rich economies (e.g., rare earths, battery metals).

### IV. Information goods, digital monopolies, and superstar effects
- Economic properties of AI-driven markets:
  - AI is an information good: non-rivalrous use and close-to-zero marginal cost enable single firms to serve very large markets.
  - High sunk/fixed costs, network externalities, and data accumulation create barriers to entry, favoring monopolies or superstar firms.
  - Superstar profits may reflect monopoly power and algorithmic extraction of consumer surplus (discriminatory pricing).
- Country-level superstar effects:
  - Countries leading in AI may reap disproportionate rents, impeding global catching-up; only a few developing countries (China noted) are exceptions.
  - Selling diverse data by developing countries is unlikely to compensate for lost income; future ML advances may reduce reliance on large data quantities.

### V. Uncertainties, scenarios, and potential magnitudes of job displacement
- Wide range of estimates on susceptibility of jobs to automation (exact phrasings preserved):
  - OECD (2019) estimate: "14% of all jobs".
  - Bain and Company (2018) estimate: 20–25%.
  - Frey and Osborne (2017) and McKinsey Global Institute (2017) estimates: almost 50%.
- Even lower percentages can have concentrated and significant distributional impacts, especially among unskilled and routine jobs.
- Extreme scenario:
  - Extensive labor-replacing innovation could render traditional labor redundant, with marginal product of human labor potentially falling below subsistence in the worst case.

### VI. Productivity puzzle and measurement issues
- Two main explanations for modest aggregate productivity growth despite transformative AI:
  - Long lags and sectoral concentration: initial AI impacts concentrated in limited sectors; complementary investments, organizational changes, and new skills take time.
  - Measurement problems: welfare gains from some digital technologies (e.g., online services financed via attention/ads) are poorly captured by GDP; improved advertising engines and market-power shifts may not show up in measured productivity.

### VII. Interaction with COVID-19, demographics, and the Green Transition
- COVID-19:
  - Accelerated automation of jobs requiring physical interaction; pandemic-introduced innovations may persist and reduce labor demand for some time.
- Demographics:
  - Rapidly growing working-age populations (many African countries) raise need for job creation; externally developed automation can make job creation harder.
  - Declining working-age populations (e.g., China) mitigate automation impact via retirements; aging societies also adopt automation.
- Green Transition:
  - Transition away from fossil fuels may redistribute demand and potentially increase labor demand (offsetting some automation effects).
  - Transition resembles resource-saving innovation that can harm fossil-fuel exporters.
  - Retraining costs for shifting workers may be moderate, but skill mismatches and institutional/finance barriers exist.
  - Policy responses include local, national, and multilateral Green Development Banks and better disclosure and fiduciary standards.

### VIII. Historical lessons and development implications
- Past technological revolutions:
  - Industrial Revolution and “Age of Labor” show labor-using revolutions can shift political dynamics in favor of workers and facilitate broad sharing of gains; labor-saving progress risks reversing these gains.
  - Pre-Industrial Revolution innovations were often absorbed by population growth (Malthusian dynamics); risk of return to Malthusian outcomes exists if developing countries lose comparative advantages.
- Implications for development strategy:
  - Manufacturing-based export-led growth (dominant successful model over past fifty years) relied on cheap labor exports and enabled structural transformation.
  - Advances in AI may erode cheap-labor advantages and make manufacturing-led export growth less viable.
  - New development strategy should be multi-pronged: leverage AI and digital tech in agriculture and services, steer adoption toward labor-using directions, encourage inward FDI aligned with labor-using objectives, and address market failures (knowledge spillovers, capital-market imperfections, mispriced externalities).

### IX. Policy recommendations — domestic toolkit
- Taxation and redistribution:
  - Capture quasi-rents from winners and redistribute to losers; shift taxation as labor tax base erodes toward other factors and rents (e.g., land, monopoly rents).
  - Pigouvian taxes on polluting goods (e.g., carbon) are desirable to fund the Green Transition.
  - Address regressive shifts in tax systems and informal-sector challenges in developing countries; digital tools may raise tax compliance.
- Social protection and UBI:
  - Social insurance to socialize risks from disruptive innovations; many developing countries have underdeveloped systems.
  - Universal Basic Income (UBI) proposals entail formidable fiscal costs; fiscal costs could be contained if UBI replaces other programs.
  - Short- to medium-run focus: create jobs for those able and willing to work, particularly in Green Transition and caregiving.
  - Alternatives/complements: shared capital ownership (e.g., sovereign wealth funds receiving firm shares).
- Expenditure and infrastructure:
  - Targeted public spending with high social returns: education, health, labor-intensive infrastructure, digital connectivity (e.g., wireless 5G), public transportation linking low-income workers to jobs, creation of service-sector jobs (healthcare, elderly care, education).

### X. Policy recommendations — pre-distribution, steering innovation, and labor-using technologies
- Pre-distribution and rules of the game:
  - Mix of pre-distribution (more equal market incomes) and redistribution is optimal.
  - Policies affecting endowments (education, financial assets) and returns on factors (competition law, labor legislation, corporate governance) shape market income distribution.
  - Equal access to high-quality education is essential but not sufficient; education alone cannot reverse declining labor income share.
- Steering innovation:
  - Direction of innovation largely set by high-income countries and China; desirable to steer technological progress toward labor-using innovations.
  - Mechanisms include nudges to entrepreneurs, aligning government-sponsored research with labor-market goals, and explicit private-sector incentives.
  - Monetary and tax policies influence automation incentives (e.g., tax policies favoring capital over labor distort direction).
  - Migration policy: restrictive migration raising wages can provide excessive incentives to automate unskilled-labor tasks.
- Location and regional policy:
  - “Good” jobs cluster in urban centers; location-based policies must weigh income growth versus geographical inequality.

### XI. Policy recommendations — supra-national reforms and global governance
- Need to update global governance to enable developing countries to maximize benefits and minimize costs of AI:

A. Global tax regime for the digital age
- Current regime allows multinationals to avoid taxation in countries where economic activity occurs.
- Formulary apportionment is proposed; formula choice (e.g., sales-only) has large distributive effects and may disadvantage developing countries.
- Initiatives include closing fiscal paradises, transparency in capital ownership, and a global minimum corporate tax rate.

B. Global competition policy
- Digital markets produce natural monopolies; competition policy must address conflicts of interest for firms that both own and participate in marketplaces, pre-emptive mergers, and ex-post remedies.
- Coordinated action among developing and emerging economies could increase leverage against global firms.

C. Intellectual Property Rights
- Current IP regime grants temporary monopoly rents; excessive protection can harm developing countries.
- Shortening patent protection or keeping innovations in the public domain (financed by governments/donors) can improve welfare sharing, especially in agriculture and healthcare.
- Differential patent protection for developing countries is justifiable; compulsory licenses (TRIPS) exist but are underused.

D. Data and information policy
- Data monopolization by global firms hinders catch-up.
- Policy responses include requiring data sharing, algorithmic and advertising transparency, strengthening privacy protections, and measures against misinformation and discriminatory targeting.
- International cooperation needed because countries hosting tech giants may lack incentives to police global behavior.

### XII. Conclusion — scenarios and policy imperatives
- Advances in AI may represent a turning point comparable to the Industrial Revolution.
- Potential adverse macro effects:
  - Increased wage inequality.
  - Declining labor demand.
  - Increased skill premia.
  - Demise of the manufacturing-export-led development model for many emerging markets.
- Worst-case scenario: unravelling of development and poverty-reduction gains of the last half century.
- Competitive equilibrium frameworks may be less applicable to the AI economy; proactive national and global policy design is warranted to mitigate large downside risks and strive for increased living standards for all.

*Source: wpiea2021166-print-pdf - chapter 1 in Cerra et al., forthcoming (IMF working paper PDF).*

### References .............................................................................................................

### wpiea2021166-print-pdf - References ...............................................................................................

### I. INTRODUCTION
- Advances in artificial intelligence (AI) and related automation technologies have led to growing fears about job losses and increasing inequality, with developing countries and emerging market economies especially vulnerable because their comparative advantage relies on abundant labor and natural resources.
- Historical context and trends:
  - For the three decades following World War II, many countries experienced broadly shared increases in living standards; over the past half-century, output growth and median worker incomes started to decouple.
  - The era of manufacturing-based export-led growth that enabled the East Asian Miracle stretched over the past half-century — characterized as only one quarter of the history of the Industrial Revolution.
- Key properties of AI highlighted:
  - AI is likely to be labor-saving and resource-saving, devaluing sources of comparative advantage for many developing countries and deteriorating their terms of trade.
  - Being an information technology, AI tends to give rise to natural monopolies, creating a small set of so-called superstar firms that are located in a few powerful countries but serve the entire world economy.
  - Under reasonable assumptions, the rate and direction of technological progress chosen by the market are generally suboptimal (Korinek and Stiglitz, 2019), creating scope to steer innovation toward socially desirable directions (e.g., preserving the planet or creating satisfying employment opportunities).
- Uncertainty and the productivity puzzle:
  - There is vast uncertainty about the impact of AI; views range from “less important than the big innovations of the 20th century” to “more rapid technological progress than mankind has ever seen before.”
  - Economic data show modest productivity increases over recent decades, encapsulated by the so-called productivity puzzle.
- Policy framing:
  - Domestic policies and international cooperation, including rewriting global rules governing the information economy, are required to mitigate adverse effects and steer technological progress toward shared prosperity.
  - Taxation, redistribution, targeted expenditure (education, infrastructure), steering technology adoption toward labor-using sectors (agriculture, services), and supra-national reforms (global tax regime for the digital age, competition policy, intellectual property, information/data policy) are presented as pillars of response.

### II. DOWNSIDE RISKS OF TECHNOLOGICAL PROGRESS
- Empirical observation:
  - The data (Figure 1) show episodes during which wages lagged productivity growth; median wages may not have kept up even where average wages did.
  - Figure 2 shows widening income inequality in most advanced economies since the early 1980s; in the US and other high-income countries, most benefits of growth have accrued to those at the top.
- Theoretical interpretation:
  - In simple competitive models with Cobb-Douglas production functions, relative shares are fixed; but in more general settings technical change can alter income distribution.
  - Hicksian taxonomy applied:
    - Capital-biased technical change: decreases the relative share of labor.
    - Skill-biased technical change: decreases the share of unskilled labor.
    - Labor-saving technical change: leads to an outright reduction in wages.
  - Korinek and Stiglitz (2019): innovations generate quasi-rents; winners (e.g., skilled workers) receive gains without contributing to innovation, losers suffer without fault. Governments can capture quasi-rents through taxation and redistribution, potentially in ways with limited distortionary effects (e.g., taxing owners of fixed factors such as land).
- Cross-border implications:
  - Distinction between within-country redistribution and global effects: when benefits accrue in one country while costs fall on another, Pareto improvement would require cross-border compensation, which today is voluntary and limited.
  - Result: some countries may gain while others lose; differences will be reflected in improvements and deteriorations of countries’ terms of trade.

### III. LABOR-SAVING TECHNOLOGICAL PROGRESS (SECTION A)
- Concerns:
  - AI may be labor-saving, causing a decline in demand for labor at existing factor prices; if so, equilibrium wages will decrease and workers will be worse off.
- Historical and empirical examples:
  - Since the 1980s the US experienced routine-biased technological change replacing workers engaged in manual and cognitive routine activities, contributing to hollowing out of the middle class (Autor et al., 2003).
  - Autor et al. (2003) observe that from the 1970s to the 1990s computerization substituted for an increasing number of routine tasks and increased productivity of workers in non-routine jobs; these changes may have explained nearly 2/3 of the relative demand shift toward college-educated labor over that period.
  - Acemoglu and Restrepo (2020) estimated significant adverse employment and wage effects from automation (study referenced; specific numerical estimates beyond those cited above are not included in the provided excerpt).

### IV. HISTORICAL LESSONS AND DEVELOPMENT IMPLICATIONS
- Historical analogy:
  - The Industrial Revolution and the ensuing Age of Labor demonstrate that labor-using technological revolutions can shift political dynamics in favor of workers and facilitate broad sharing of gains; however, labor-saving progress risks reversing such gains.
- Development strategy implications:
  - Decline of manufacturing requires a multi-pronged development model to replace manufacturing-based export-led growth.
  - Developing countries should leverage AI and digital technologies in agriculture and services and steer technology adoption to create labor demand and preserve comparative advantages where possible.

### V. POLICY RECOMMENDATIONS (DOMESTIC AND SUPRA-NATIONAL)
- Domestic policy toolkit:
  - Taxation and redistribution:
    - Capture quasi-rents from winners of technological progress and redistribute to losers; scope may be limited in developing countries but remains a first line of defense.
  - Targeted expenditure:
    - Invest in education and infrastructure that are labor intensive and enhance human capital and physical infrastructure to bridge the digital divide and enable participation in digital opportunities.
  - Steering technology adoption:
    - Promote labor-using directions of technological progress and leverage AI in sectors where developing countries have potential comparative advantages (e.g., agriculture, services).
- Supra-national reforms:
  - Global tax regime for the digital age:
    - Enable countries to raise taxes on transactions occurring within their borders to address tax base erosion by digital firms.
  - Competition policy:
    - Address global footprint of digital giants; national authorities in origin countries may have incorrect incentives to ensure competitive markets globally.
  - Intellectual property regimes:
    - Adapt IP rules to the needs and circumstances of developing countries.
  - Information and data policy:
    - Discuss regulation of data at the supra-national level to give developing countries a voice in policy design.

### VI. ORGANIZATION OF THE PAPER (AS OUTLINED)
- Subsequent sections of the paper (as described in the excerpt) cover:
  - An overview of downside risks of technological progress, with emphasis on AI-induced disruptions.
  - Discussion of uncertainties surrounding the nature and level of impacts and broader context (COVID-19 recovery, population dynamics, Green Transition).
  - Review of historical perspectives on technological progress.
  - Distillation of the role of government policy in managing effects and enabling shared benefits.
  - Analysis of how global governance needs updating to allow developing countries to maximize benefits and minimize costs of AI and related technologies.

*Source: wpiea2021166-print-pdf - References .....................................................................................................*

### introduction of industrial robots in the US, concentrated in manufacturing and among routine

### wpiea2021166-print-pdf - introduction of industrial robots in the US, concentrated in manufacturing and among routine

### Job polarization and wage-skill premia
- Real wages of full-time U.S. male workers indexed to 1963 = 100 show rising wage-skill premia across educational groups (Figure 3).
- Employment dynamics:
  - Employment in nonroutine jobs has continued to grow steadily in the US.
  - Employment in routine jobs has stagnated, or in some periods declined, since around 1990, contributing to a “hollowing out of the middle” (Figure 4).
- COVID-19 (2020) accelerated the trend temporarily, producing a large decline in employment in routine manual jobs but only a modest dip in nonroutine cognitive jobs.

### Mechanisms and theoretical models
- Standard aggregate production function frameworks with skilled and unskilled labor-augmenting progress and capital-augmenting progress can generate observed factor price and share movements depending on progress patterns and elasticities.
- Acemoglu and Restrepo (2019a): displacement of workers by robots can reduce the labor share of income and may be labor-saving if productivity gains from robots are modest.
- Berg et al. (2018): technological progress may be unskilled-labor-saving (robots substitute unskilled labor) while high-skilled labor is complementary to robots, risking large increases in inequality.
- Long-run effects may differ: additional capital accumulation complementary to labor could benefit labor (Stiglitz, 2015; Caselli and Manning, 2019), but if other scarce factors (e.g., natural resources or land) become limiting, labor may lose without government intervention (Korinek and Stiglitz, 2021a).

### Distributional and global impacts
- Labor-saving technological progress:
  - Makes the world richer overall but can disproportionately harm developing countries with a comparative advantage in cheap labor.
  - Could cause deterioration in terms of trade and large export-income losses for such countries.
  - Alonso et al. (2020): improvements in robot productivity could drive divergence as advanced countries benefit more given higher initial capital stocks.
- Resource-saving technological progress:
  - AI and digital technologies can reduce demand for depletable natural resources and lower carbon emissions (examples: efficiency algorithms for data centers, transportation; telework reducing carbon footprint).
  - Adverse distributional effects for countries with comparative advantage in natural resources (e.g., oil-exporters). Some resource-rich economies may benefit (rare earths, metals for batteries, microchips, solar panels, wind turbines).
  - For oil-exporting countries, loss of oil export revenue could be dire for food and essentials imports.

### Task-level effects, AI vs. IA, and job restructuring
- Automation often affects particular tasks, not entire jobs; jobs consist of multiple tasks.
- Intelligence-assisting devices and algorithms (IA) may be complementary to labor, improving productivity by taking over or improving certain tasks (examples: AI in diagnosis vs. doctor-patient communication; partial automation in truck driving).
- Both AI and IA imply extensive restructuring of the economy and reallocation of worker attention toward non-automated tasks.

### Information, digital monopolies, and superstar effects
- AI is an information good: non-rivalrous use and close-to-zero marginal cost can enable single firms to serve very large markets.
- High sunk/fixed costs, network externalities, and data accumulation create large barriers to entry and a tendency toward monopolies or “superstar” firms.
- Monopoly power may drive much of the profits of superstar firms rather than purely super-productive technology; algorithmic advances enable extraction of consumer surplus (discriminatory pricing).
- Superstar effects can arise at the country level: countries leading in AI may reap disproportionate rents, potentially impeding global catching-up; China noted as a notable exception among developing countries.
- Selling diverse data by developing countries is unlikely to compensate for lost income; future ML advances may reduce reliance on large data quantities.

### Misguided technological progress and induced-innovation considerations
- Direction and rate of technological progress are economic decisions; markets do not generally choose innovation directions that are welfare-maximizing.
- Knowledge is a Samuelsonian public good; private provision without restraints leads to either inefficient appropriation (market power) or underinvestment (if no appropriation).
- Market biases toward innovations with high private returns (e.g., easily secured intellectual property, me-too drugs) can misalign social and private incentives.
- Pecuniary externalities and incomplete markets can bias innovation toward labor-augmenting (and ultimately unemployment-increasing) directions in some models (e.g., Shapiro-Stiglitz efficiency wage).
- Markets do not internalize distributive consequences; this creates a rationale for steering innovation and for redistributive policies when warranted.

### Broader harms associated with AI
- Security (including cybersecurity), privacy, incitement (hate speech), political manipulation, and economic harms (price discrimination) are notable concerns.
- International regulatory and standards-setting discussions may be dominated by high-income countries and China, potentially not reflecting developing-country priorities.
- Institutional capacity constraints in developing countries increase vulnerability to mis-/disinformation and surveillance abuses.

### Uncertainties, scenarios, and potential magnitudes of job displacement
- Uncertainty about pace and scale of AI-driven change is large; slow, task-limited progress allows adjustment whereas rapid, economy-wide change is disruptive.
- Views range widely:
  - OECD (2019) estimate: 14% of all jobs (preservation of exact phrasing from source: "14% of all jobs").
  - Bain and Company (2018) estimate: 20–25%.
  - Frey and Osborne (2017) and McKinsey Global Institute (2017) estimates: almost 50%.
- Even lower percentages can have concentrated and significant distributional impacts, especially among unskilled and routine jobs.
- Extreme scenario: extensive labor-replacing innovation could render traditional labor redundant, with marginal product of human labor potentially falling below subsistence (Korinek and Stiglitz, 2019).

### Productivity puzzle and measurement issues
- If AI is transformative, why is GDP growth not reflecting it? Two main explanations:
  - Long lags and sectoral concentration (initially AI affects limited sectors; complementary investments, organizational changes, and new skills needed).
  - Measurement problems: welfare gains from some technologies (e.g., online services paid via attention/ads) are poorly captured by GDP; better advertising engines may alter market power and efficiency.

### Interaction with COVID-19, demographics, and the Green Transition
- COVID-19 accelerated automation of jobs requiring physical interaction; innovations introduced during the pandemic may persist and reduce labor demand worldwide for some time.
- Population dynamics:
  - Rapidly growing working-age populations (many African countries) increase need for job creation; automation developed externally can make this harder.
  - Declining working-age populations (e.g., China) mitigate automation impact via retirements; aging societies adopt new technologies and automate (Acemoglu and Restrepo, 2019b).
- Green Transition:
  - Transition away from fossil fuels can redistribute demand and may increase labor demand (offsetting some automation effects), but also resembles resource-saving innovation that can harm fossil-fuel exporters.
  - Complementarity: Green Transition may increase labor demand; retraining costs for shifting workers may be moderate (Louie and Pearce, 2016), though skill mismatches exist.
  - Institutional and finance barriers can impede green investment; local, national, and multilateral Green Development Banks and better disclosure and fiduciary standards could help.

### Lessons from past technological transformations
- Pre-Industrial Revolution: innovations were absorbed by population growth (Malthusian dynamics); risk of return to Malthusian outcomes exists if developing countries lose comparative advantages.
- Industrial Revolution:
  - Innovation can be highly disruptive in the short run; not a Pareto improvement initially.
  - Collective action and government interventions (sanitation, public health, education, labor legislation, social safety nets) mitigated adverse effects over time.
  - Labor protections and institutions underpinned the “Age of Labor”; erosion of such protections (e.g., declining real minimum wages in recent decades in the US) and AI advances may further weaken labor’s bargaining position.

*Source: wpiea2021166-print-pdf - introduction of industrial robots in the US, concentrated in manufacturing and among routine*

### chapter 1 in Cerra et al., forthcoming). But in recent decades, in many countries, the political

### chapter 1 in Cerra et al., forthcoming

### C. Manufacturing-Based Export-Led Growth — findings
- The dominant successful development model in developing countries over the past fifty years has been manufacturing-based export-led growth.
- This model moved policy from static comparative advantage to dynamic comparative advantage, exemplified by South Korea’s industrial policy and subsequent industrialization.
- The East Asia development path relied on exports of cheap labor-intensive manufactured goods, delivering employment opportunities, foreign exchange, tax revenue, and rapid developmental transition.
- Over time many East Asian countries moved up the value chain to higher value added and more complex products, developing dynamic comparative advantage.
- Earlier technology reduced the importance of cheap labor; advances in AI may erode it further, making growth led solely by exports of labor-intensive manufacturing goods no longer a viable development strategy.
- Globally:
  - The share of manufacturing employment is decreasing.
  - Jobs that can be outsourced may be more easily automated, and reshoring using highly automated production processes may accelerate (potentially hastened by the Covid pandemic).
- Technology adoption lags have declined historically (Comin and Hobijn, 2010), but the specific nature of AI—specialized, driven by large datasets, with a winner-takes-all dynamic—may reverse that trend.
- Knowledge production for cutting-edge AI is concentrated (private companies, proprietary data), making it harder for developing and emerging economies to access and catch up; the exponential growth of AI may expand gaps between front runners and laggards.
- The emphasis in export-led growth is on traded goods rather than non-traded services; non-traded service sector development on its own typically has not had a substantial impact on overall productivity.

### D. What Is Different This Time — risks and distributional effects
- Structural transformations linked to earlier industrialization tended to be unskilled-biased and egalitarian: industrial production increased relative productivity of unskilled labor, promoted mass education, and easier unionization supporting wage compression.
- The AI-driven transition risks producing a service-sector dominated economy marked by:
  - Greater inequality.
  - Less support for public education.
  - Greater concentration of market power.
  - More disruptive adjustment processes unless countervailing policies are implemented.
- AI characteristics:
  - Likely labor-saving and resource-saving.
  - Biased towards ever-higher skills; general education may become less important.
  - Creates barriers to entry and monopoly power with winner-takes-all rents accruing disproportionately to a small number of extremely wealthy individuals and enterprises, often in high-income countries.
- Potential outcomes:
  - Many developing countries may see average income per capita increase while large fractions of society are left behind.
  - Some developing countries may experience declines in income per capita as innovation erodes their comparative advantage; unskilled workers may suffer the most.
  - Concentrations of economic and political power may reduce support for government mitigation of adverse distributional consequences.

### V. Domestic policy responses — taxation, redistribution, and government expenditures
- Progressive taxation and redistribution are critical to combat rising inequality, but many countries have made tax systems more regressive; returns to capital and rents often taxed at lower rates than labor.
- Examples and evidence:
  - Based on data for 1981-2002, Gilens (2005) finds US policy outcomes reflected preferences of higher-income groups with little relationship to preferences of the poor or middle-income citizens.
  - In the US, the rich pay a lower fraction of their income in taxes than the majority of the population (Saez and Zucman, 2019).
- Challenges for developing countries:
  - Large informal sectors make raising taxes difficult but also present scope to enhance tax structures and expenditure systems (e.g., scrapping harmful subsidies and tax exemptions).
  - New digital tools and centralized digital platforms may increase tax compliance by making business transactions more accessible to governments.
- Shifts in tax bases are necessary as labor-saving progress reduces labor tax revenue; taxation should increasingly target other factors and rents (e.g., land, monopoly rents, rents from political activity).
- Pigouvian taxes on “bads” (e.g., polluting or carbon-emitting goods) are desirable for efficiency and to fund the Green Transition.
- Tax incidence concerns require attention: e.g., capital taxation may discourage accumulation, but adverse effects can be offset by public investments in human and physical capital.
- Some monopoly rents of digital giants can be taxed without major distortions.

Social protection and UBI
- Social insurance can socialize risks from disruptive innovations; systems are typically less developed in developing countries, leaving groups worse off.
- Universal Basic Income (UBI):
  - Proposals vary; typically entail unconditional payments above the poverty line.
  - UBI implies formidable fiscal costs and potential distortionary taxes, but fiscal costs could be contained if UBI replaces other social safety programs.
  - A global UBI covering all citizens equally is desirable from a global welfare perspective but is unrealistic given cross-border transfer limitations.
  - Short- to medium-run focus should be on creating jobs for everyone able and willing to work, particularly for tasks needed in the Green Transition and caregiving.
  - Alternatives or complements to UBI include shared capital ownership (e.g., firms receiving government help contributing shares to a sovereign wealth fund owned by citizens).
- Historical and behavioral notes:
  - Keynes (1931) and Varian (2020) argued automation could enable people to work less if wages rise with productivity or gains are shared widely.
  - Dutch model example: right to part-time work at pro-rated wages could serve as a model if wages are sufficiently high.

Expenditure and infrastructure policy
- Government expenditures can be easier to target to those with high social returns (education, health, environmental protection).
- Expenditure policies that increase demand for unskilled labor serve dual goals: raising pre-distribution equality and targeting benefits to the less well-off (e.g., labor-intensive infrastructure investments in poorer neighborhoods).
- Investments in digital infrastructure can reduce the “digital divide”; recent network technology advances allow leapfrogging older technologies (e.g., wireless 5G).
- Public transportation investments that connect lower income workers to jobs and creation of service-sector jobs (healthcare, elderly care, aspects of education) can increase labor demand and disproportionately benefit the poor.

### V.B. Pre-distribution and steering market income distribution
- A mix of pre-distribution (increasing equality of market income) and redistribution is optimal; many pre-distribution measures also enhance efficiency.
- Two policy categories affecting market income distribution:
  1. Policies affecting individuals' endowments of assets: human capital (education) and financial assets; public provision of education and policies affecting intergenerational transmission (e.g., inheritance taxes).
  2. Policies affecting returns on factors: laws and regulations shaping “rules of the game” (competition law, labor legislation, globalization rules, financial sector regulation, corporate governance).
- Education:
  - Equal access to high-quality education is essential to prevent an education-based digital divide, but education alone is not a panacea and can amplify innate ability differences, potentially increasing market income inequality.
  - Education cannot address the declining share of labor income overall.

Steering innovation and labor-using technologies
- Direction of innovation is largely set by high-income countries plus China; their choices (how labor-saving innovations are) affect developing countries exposed to these technologies.
- Korinek and Stiglitz (2020): desirable to steer technological progress toward labor-using innovations when lump-sum transfers are unavailable; methods include nudging entrepreneurs, considering labor-market implications of government-sponsored research, and providing explicit incentives to the private sector.
- Governmental policies indirectly influence innovation incentives (e.g., monetary policy affecting real returns on safe assets and thus automation incentives; tax policies favoring capital over labor distort direction of progress).
- Migration policy effects: restrictive migration in high-income countries raises wages, providing excessive incentives to automate tasks performed by unskilled labor (Pritchett, 2019).
- Regional heterogeneity matters: “good” jobs may cluster in major urban centers, suggesting need for location-based development policies with trade-offs between income growth and geographical inequality.

New development strategies for developing countries
- Manufacturing-led export-based growth needs to be replaced by multi-pronged strategies extending focus beyond manufacturing to agriculture and services.
- Sectoral development policies (infrastructure, education, tax, regulation) shape growth and can be subject to capture by special interests.
- Developing countries should focus on steering adoption of labor-using technologies developed elsewhere: adapting, redesigning, and building on them; encourage inward FDI aligned with these objectives.
- Market failures justify public intervention: knowledge spillovers, risk and capital market imperfections, and mispriced externalities (e.g., zero carbon price) mean firms acting alone under-invest in socially desirable directions.
- AI-specific opportunities for developing countries:
  - Agriculture: AI algorithms can increase productivity but need customization to local crops, soil, and weather; extension services have an important role.
  - Digital platforms can reduce middlemen power and improve small farmers’ market access.
  - Services: carve out comparative advantage in services dependent on internet connectivity and workforce education (e.g., call centers, business services); some services are more automation-resistant (e.g., tourism).
  - Domestic services (healthcare, elderly care, education) are crucial for welfare and may benefit from government-supported AI investments; opportunities exist for cross-border trade in services (medical tourism, retirees relocating).

### VI. Global governance — areas for international reform
- Global rules have disproportionately favored high-income countries and are set largely by powerful countries and interests; these rules affect developing countries’ ability to tax digital activity, secure fair terms of trade, and access technology.
- AI and information rules are currently being set; this presents both risks and opportunities for more inclusive governance if international institutions and civil society engage effectively.
- Key global governance reforms that would help developing countries adapt to AI:

A. Global tax regime for the digital age
- Current global tax regime allows multinationals to avoid taxation and impairs developing countries’ ability to tax activity within their territories.
- Debates at the OECD aim to establish a global tax regime.
- Formulary apportionment (apportioning worldwide profits via a formula) is a proposed remedy; the formula choice (e.g., sales-only) can have large distributive effects and may disadvantage developing countries.
- Controversy exists over how to value data collected in transactions and whether that value should be taxed.
- Other initiatives include closing fiscal paradises, transparency in capital ownership, and a global minimum corporate tax rate to prevent a race to the bottom.

B. Global competition policy
- Digital technologies tend to produce natural monopolies; competition policy is crucial.
- Countries hosting tech giants may have incentives to protect domestic firms because they share in global rents (examples: EU investigations of Google and Germany’s privacy probes of Facebook drew US political responses).
- Individual developing countries have limited power to police global corporations; coordinated action (e.g., a common competition authority for developing and emerging economies) could increase leverage.
- Policy needs include preventing conflicts of interest for companies that both own and participate in marketplaces, preventing pre-emptive mergers, and enabling ex-post remediation (breaking up mergers proven anti-competitive).

C. Intellectual Property Rights
- Current IP regime grants temporary monopoly rents to innovators; concerns exist that protection is excessive and harms developing countries.
- Reducing patent protection length can improve welfare sharing of AI-based innovations (Korinek and Stiglitz, 2019).
- Keeping innovations in the public domain, financed by governments or donors, avoids monopolies and restricted access; this is especially beneficial for agriculture and healthcare innovations relevant to developing countries.
- Differential patent protection in developing countries is justifiable; most patents are developed in high-income countries and are financed by surplus extracted there.
- Compulsory licenses exist (TRIPS) but many developing countries lack capacity or are intimidated from using them.
- AI algorithms may not need patent protection but can be kept proprietary; requiring disclosure of key algorithms is important to detect discriminatory behavior.

D. Data and information policy
- Data is a critical input for AI; monopolization of data by global firms hinders developing countries’ ability to catch up.
- Global tech firms are influencing data regulation via trade agreements, often favoring firm interests over public interest.
- Policy responses:
  - Require data sharing to prevent monopoly power accretion (EU proposals).
  - Recognize that giving data control rights to individuals alone is insufficient—unequal power and information asymmetries hamper equitable outcomes.
  - Enact transparency regulations on algorithms and advertising targeting, and address discriminatory pricing and targeting.
  - Strengthen privacy protections and measures against misinformation, violent or hateful messaging, and reconsider provisions like Section 230 that reduce platform accountability.
- Developing countries need to cooperate internationally to impose regulations that reflect developmental interests because countries hosting tech giants may lack incentives to police global behavior.

### VII. Conclusion — scenarios and implications
- Advances in AI and related technologies may represent a turning point comparable to the Industrial Revolution.
- Potential adverse macro effects include:
  - Increased wage inequality.
  - Declining labor demand.
  - Increased skill premia.
  - Demise of the manufacturing-export-led development model that aided many emerging markets.
- Worst-case scenario: unravelling of gains in development and poverty reduction observed over the last half century.
- Unlike earlier technological advances that were associated with shared prosperity, current advances risk increasing inequality within and between countries unless counteracted by policy.
- The new era requires different economic models and analysis; competitive equilibrium frameworks may be less applicable to the 21st century AI economy.
- Given uncertainty across scenarios, there are large potential downside risks that warrant proactive policy design at national and global levels to mitigate adverse effects and strive for increased living standards for all, including billions in developing countries.

*Italic — source: wpiea2021166-print-pdf - chapter 1 in Cerra et al., forthcoming (IMF working paper PDF).*

### REFERENCES

### REFERENCES

### Directed technical change, automation, and labor markets
- Acemoglu, Daron (1998), “Why Do New Technologies Complement Skills? Directed Technical Change and Wage Inequality,” Quarterly Journal of Economics, 113(4), pp. 1055-1089.  
- Acemoglu, Daron (2002), “Directed Technical Change,” Review of Economic Studies, 69(4), pp. 781-809.  
- Acemoglu, Daron and David Autor (2011), “Skills, Tasks and Technologies: Implications for Employment and Earnings,” Handbook of Labor Economics 4b, pp. 1043-1171.  
- Acemoglu, Daron, Andrea Manera, and Pascual Restrepo (2020), “Does the US tax code favor automation?” Brookings Papers on Economic Activity 2020(1).  
- Acemoglu, Daron, and Pascual Restrepo (2019a), “Automation and New Tasks: How Technology Displaces and Reinstates Labor,” Journal of Economic Perspectives 33(2), pp. 3–30.  
- Acemoglu, Daron, and Pascual Restrepo (2019b), “Demographics and Automation,” Unpublished Working Paper, MIT. https://economics.mit.edu/files/16788  
- Acemoglu, Daron, and Pascual Restrepo (2020), “Robots and Jobs: Evidence from US Labor Markets,” Journal of Political Economy 128(6), pp. 2188–2244.  
- Autor, David H., (2014), “Skills, Education, and the Rise of Earnings Inequality Among the ‘Other 99 Percent’,” Science, 23 May 2014: 344 (6186), pp. 843-851.  
- Autor, David (2015), “Why Are There Still So Many Jobs? The History and Future of Workplace Automation,” Journal of Economic Perspectives 29(3), pp. 3–30.  
- Autor, David H. (2019), “Work of the Past, Work of the Future,” AEA Papers and Proceedings 109, pp. 1-32.  
- Autor, David, David Dorn, Lawrence F Katz, Christina Patterson, and John Van Reenen (2020), “The Fall of the Labor Share and the Rise of Superstar Firms,” Quarterly Journal of Economics 135(2), pp. 645-709.  
- Autor, David, Frank Levy, and Richard J. Murnane, (2003), “The Skill Content of Recent Technological Change: An Empirical Exploration,” Quarterly Journal of Economics 118(4), pp. 1279-1333.  
- Caselli, Francesco, and Alan Manning (2019), “Robot Arithmetic: New Technology and Wages,” American Economic Review: Insights 2019 1(1), pp. 1-12.  
- Frey, Carl Benedikt and Michael A. Osborne (2017), “The future of employment: How susceptible are jobs to computerisation?” Technological Forecasting and Social Change 114, pp. 254-280.  
- Korinek, Anton (2019), “The Rise of Artificially Intelligent Agents,” working paper, University of Virginia.  
- Korinek, Anton (2020), “Taxation and the Vanishing Labor Market in the Age of AI,” Ohio State Technology Law Journal 16(1), pp. 244-257.  
- Korinek, Anton and Ding Xuan Ng (2019), “Digitization and the Macro-Economics of Superstars,” working paper, University of Virginia.  
- Korinek, Anton and Joseph E. Stiglitz (2019), “Artificial Intelligence and Its Implications for Income Distribution and Unemployment,” in Agrawal et al.: The Economics of Artificial Intelligence, NBER and University of Chicago Press.  
- Korinek, Anton and Joseph E. Stiglitz (2020), “Steering Technological Progress,” working paper.  
- Korinek, Anton and Joseph E. Stiglitz (2021a), “Artificial Intelligence, Globalization, and Strategies for Economic Development,” NBER Working Paper w28453.  
- Korinek, Anton and Joseph E. Stiglitz (2021b), “Will COVID-19 Drive Advances in Automation and AI that Exacerbate Economic Inequality?” BMJ 372:n367.  
- Peralta-Alva, Adrian, and Agustin Roitman (2018), “Technology and the Future of Work,” IMF Working Paper WP/18/207. Also published as IMF Staff Note to Group of Twenty: April 2018.  
- Summers, Lawrence (2017), “Picking on robots won’t deal with job destruction,” The Washington Post Opinion, March 5, 2017.  
- Varian, Hal (2020), “Automation versus procreation (aka bots versus tots),” VoxEU, https://voxeu.org/article/automation-versus-procreation-aka-bots-versus-tots

### Artificial intelligence, predictions, and the singularity debate
- Aghion, Philippe, Benjamin F. Jones & Charles I. Jones (2017), “Artificial Intelligence and Economic Growth,” NBER WP 23928.  
- Armstrong, Stuart, Kaj Sotala and Seán S. ÓhÉigeartaigh (2014), “The errors, insights and lessons of famous AI predictions – and what they mean for the future,” Journal of Experimental & Theoretical Artificial Intelligence, 26(3), pp. 317-342. http://www.fhi.ox.ac.uk/wp-content/uploads/FAIC.pdf  
- Grace, Katja, John Salvatier, Allan Dafoe, Baobao Zhang, and Owain Evans (2018), “When Will AI Exceed Human Performance? Evidence from AI Experts,” Journal of Artificial Intelligence Research, https://arxiv.org/abs/1705.08807v3.  
- Kurzweil, Ray (2005) The Singularity is Near, Penguin Group, pp. 135–136.  
- Ulam, Stanislaw (1958), "Tribute to John von Neumann," Bulletin of the American Mathematical Society 64, #3, part 2, pp. 5.  
- Vinge, Vernor (1993), "The Coming Technological Singularity: How to Survive in the Post-Human Era,” Vision-21: Interdisciplinary Science and Engineering in the Era of Cyberspace, Geoffrey Landis, ed., NASA Publication CP-10129, pp. 11–22. https://edoras.sdsu.edu/~vinge/misc/singularity.html  
- Walsh, Toby (2016), “The Singularity May Never Be Near,” https://arxiv.org/abs/1602.06462v1.  
- Good, Irving John (1965), “Speculations Concerning the First Ultraintelligent Machine,” Advances in Computers, vol. 6.

### Inequality, inclusive growth, and policy responses
- Aghion, Philippe, Reda Cherif, and Fuad Hasanov (2021), “Competition, Innovation, and Inclusive Growth,” IMF Working Paper No. 2021/080. Also forthcoming as Chapter 6 in Cerra et al. (forthcoming).  
- Aghion, Philippe, Ufuk Akcigit, Antonin Bergeaud, Richard Blundell, and David Hemous (2019), “Innovation and Top Income Inequality,” Review of Economic Studies 86(1), pp. 1-45.  
- Alonso, Cristian, Andrew Berg, Siddharth Kothari, Chris Papageorgiou, and Sidra Rehman (2020), “Will the AI Revolution Cause a Great Divergence?,” IMF Working Paper 20/184.  
- Brussevich, Mariya, Era Dabla-Norris, Christine Kamunge, Pooja Karnane, Salma Khalid, and Kalpana Kochhar (2018), “Gender, Technology, and the Future of Work,” IMF Staff Discussion Note, SDN/18/07, October 2018.  
- El-Ganainy, Asmaa, Ekkehard Ernst, Rossana Merola, Richard Rogerson, and Martin Schindler (2021), “Inclusivity in the Labor Market,” IMF Working Paper No. 2021/141. Also forthcoming as Chapter 3 in Cerra et al. (forthcoming).  
- Klinova, Katya and Anton Korinek (2021), “AI and Shared Prosperity,” Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society (AIES ’21).  
- Partnership on AI (2021), Redesigning AI for Shared Prosperity: An Agenda. Technical Report. Available at: partnershiponai.org/shared-prosperity  
- Ostry, Jonathan, Prakash Loungani, and Andrew Berg (2019), Confronting Inequality: How Societies Can Choose Inclusive Growth, Columbia University Press.  
- Peterson Institute for International Economics (PIIE) (2020), “How to Fix Economic Inequality?” https://www.piie.com/sites/default/files/documents/how-to-fix-economic-inequality.pdf.  
- Pouokam, Nathalie (2021), “Sharing Resource Wealth and Addressing Fragility,” IMF Working Paper No. 2021/097. Also forthcoming as Chapter 6 in Cerra et al. (forthcoming).  
- Saez, Emmanuel and Gabriel Zucman (2019), The Triumph of Injustice: How the Rich Dodge Taxes and How to Make Them Pay, Norton.  
- Stiglitz, Joseph E. (2013) The Price of Inequality: How Today’s Divided Society Endangers Our Future, New York: W.W. Norton & Company.  
- Stiglitz, Joseph E., Nell Abernathy, Adam Hersh, Susan Holmberg and Mike Konczal (2015), Rewriting the Rules of the American Economy: An Agenda for Growth and Shared Prosperity, A Roosevelt Institute Book, Norton.  
- Stiglitz, Joseph E., Carter Daugherty, and the Foundation for European Progressive Studies (2019), Rewriting the Rules of the European Economy: An Agenda for Growth and Shared Prosperity, Norton.  
- Rubin, Richard (2020), “The ‘Robot Tax’ Debate Heats Up,” Wall Street Journal, January 20, 2020. https://www.wsj.com/articles/the-robot-tax-debate-heats-up-11578495608.  
- Clausing, Kimberley and Reuben Avi-Yonah (2007), “Reforming Corporate Taxation in a Global Economy: A Proposal to Adopt Formulary Apportionment,” The Hamilton Project, Brookings Institution.

### Innovation, growth theory, learning, and technological diffusion
- Arrow, Kenneth (1962a), “Economic Welfare and the Allocation of Resources for Invention,” in: R. Nelson, ed., The Rate and Direction of Inventive Activity: Economic and Social Factors, Princeton University Press.  
- Arrow, Kenneth (1962b), “The Economic Implications of Learning by Doing,” The Review of Economic Studies 29(3), pp. 155-173.  
- Atkinson, Anthony, and Joseph Stiglitz (1969), “A New View of Technological Change,” Economic Journal 79 (315), pp. 573-578.  
- Kennedy, Charles (1964), “Induced Bias in Innovation and the Theory of Distribution,” Economic Journal LXXIV, pp. 541– 47.  
- Romer, Paul M. (1986), “Increasing Returns and Long-Run Growth,” Journal of Political Economy, Vol. 94, No. 5 (October), pp. 1002-1037.  
- Solow, Robert (1987), “We’d Better Watch Out.” New York Times Book Review, July 12, pp. 36.  
- Solow, Robert (2009), “Does Growth Have a Future? Does Growth Theory Have a Future? Are These Questions Related?,” History of Political Economy 41, pp. 27-34.  
- Comin, Diego, and Bart Hobijn (2010), “An Exploration of Technology Diffusion,” American Economic Review 100(5), pp. 2031-59.  
- Maddison, Angus (2003), Contours of the World Economy 1-2030 AD: Essays in Macro-Economic History, Oxford University Press.  
- North, (not listed) — (Note: only the supplied entries above are included.)

### Industrial policy, development, and structural change
- Cimoli, Mario, Giovanni Dosi, Keith E. Maskus, Ruth L. Okediji, Jerome H. Reichman, and Joseph Stiglitz (eds.) (2014), Intellectual Property Rights: Legal and Economic Challenges for Development, Oxford University Press.  
- Delli Gatti, Domenico, Mauro Gallegati, Bruce Greenwald, Alberto Russo, and Joseph E. Stiglitz (2012a), “Sectoral Imbalances and Long Run Crises,”The Global Macro Economy and Finance, Franklin Allen, Masahiko Aoki, Jean-Paul Fitoussi, Nobuhiro Kiyotaki, Roger Gordon, and Joseph Stiglitz, eds., IEA Conference Volume No. 150-III, Palgrave, pp. 61-97.  
- Delli Gatti, Domenico, Mauro Gallegati, Bruce Greenwald, Alberto Russo, and Joseph E. Stiglitz (2012b), “Mobility Constraints, Productivity Trends, and Extended Crises,” Journal of Economic Behavior & Organization 83(3), pp. 375– 393.  
- Greenwald, Bruce and Joseph Stiglitz (1986) “Externalities in Economies with Imperfect Information and Incomplete Markets,” Quarterly Journal of Economics, Vol. 101, No. 2, May, pp. 229-264.  
- Greenwald, Bruce and Joseph Stiglitz (2006) “Helping Infant Economies Grow: Foundations of Trade Policies for Developing Countries,” American Economic Review: AEA Papers and Proceedings 96(2), pp. 141-146.  
- Greenwald, Bruce and Joseph Stiglitz (2014a), Creating a Learning Society: A New Approach to Growth, Development, and Social Progress, Columbia University Press.  
- Greenwald, Bruce and Joseph E. Stiglitz (2014b), “Industrial Policies, the Creation of a Learning Society, and Economic Development,” The Industrial Policy Revolution I: The Role of Government Beyond Ideology, Joseph E. Stiglitz and Justin Yifu Lin (eds.), Palgrave Macmillan, pp. 43-71.  
- McMillan, Margaret S., and Dani Rodrik (2011), “Globalization, Structural Change and Productivity Growth,” NBER Working Paper 17143.  
- Rodrik, Dani (2011), “Unconditional Convergence,” NBER Working Paper w17546.  
- Pritchett, Lant (2019), “The Future of Jobs is Facing One, Maybe Two, of the Biggest Price Distortions Ever,” Economic Research Forum Working Paper 1370. https://erf.org.eg/app/uploads/2019/12/1370.pdf

### Firm dynamics, market power, and antitrust
- Akcigit, Ufuk, Wenjie Chen, Federico J. Díez, Romain Duval, Philipp Engler, Jiayue Fan, Chiara Maggi, Marina M. Tavares, Daniel Schwarz, Ippei Shibata, and Carolina Villegas-Sánchez (2021), “Rising Corporate Market Power: Emerging Policy Issues,” IMF Staff Discussion Note SDN/21/01.  
- Autor, David, David Dorn, Lawrence F Katz, Christina Patterson, and John Van Reenen (2020), “The Fall of the Labor Share and the Rise of Superstar Firms,” Quarterly Journal of Economics 135(2), pp. 645-709.  
- Choi, Eleanor J., and James R. Spletzer (2012), “The declining average size of establishments: evidence and explanations,” BLS Monthly Labor Review, March 2012. https://www.bls.gov/opub/mlr/2012/03/art4full.pdf  
- Wu, Tim (2018), The Curse of Bigness: Antitrust in the New Gilded Age, Columbia Global Reports.

### Environment, decarbonization, and pollution disparities
- Colmer, Jonathan, Ian Hardman, Jay Shimshack, and John Voorheis (2020), “Disparities in PM2.5 air pollution in the United States,” Science, Vol. 369, Issue 6503, pp. 575-578.  
- Peszko, Grzegorz; van der Mensbrugghe, Dominique; Golub, Alexander; Ward, John; Zenghelis, Dimitri; Marijs, Cor; Schopp, Anne; Rogers, John A.; Midgley, Amelia (2020). Diversification and Cooperation in a Decarbonizing World: Climate Strategies for Fossil Fuel-Dependent Countries. Washington, DC: World Bank.  
- Louie, Edward P., and Joshua M. Pearce (2016), “Retraining investment for U.S. transition from coal to solar photovoltaic employment,” Energy Economics 57, pp. 295-302.

### Historical, philosophical, and broader perspectives
- George, Henry (1879), Progress and poverty: An inquiry into the cause of industrial depressions, and of increase of want with increase of wealth, the remedy, D Appleton and Company.  
- Malthus, Thomas R. (1798), An Essay on the Principle of Population, J. Johnson.  
- Myrdal, Gunnar (1968), Asian Drama: An Inquiry into the Poverty of Nations, Pantheon Books.  
- Gordon, Robert (2016), The Rise and Fall of American Growth: The U.S. Standard of Living since the Civil War, Princeton University Press.  
- Harari, Yuval Noah (2017), Homo Deus: A Brief History of Tomorrow, Harper.  
- Trajtenberg, Manuel (2019) “Artificial Intelligence as the Next GPT: A Political-Economy Perspective,” pp. 175-186, in Agrawal, Ajay, Joshua Gans and Avi Goldfarb (2019), The Economics of Artificial Intelligence: An Agenda, National Bureau of Economic Research and University of Chicago Press.

### Major institutional reports and datasets
- McKinsey Global Institute (2017), “Jobs lost, jobs gained: What the future of work will mean for jobs, skills, and wages,” McKinsey Reports.  
- OECD (2019), The Future of Work, OECD Employment Outlook 2019.  
- World Bank (1998), World Development Report 1998/1999: Knowledge for Development, Oxford University Press. https://openknowledge.worldbank.org/handle/10986/5981  
- World Commission on the Social Dimension of Globalization (2004), A Fair Globalization: Creating Opportunities for All, https://www.ilo.org/public/english/wcsdg/docs/report.pdf  
- World Economic Outlook (WEO) (2018), “Manufacturing Jobs: Implications for Productivity,” Chapter 3, April 2018.

*REFERENCES — wpiea2021166-print-pdf*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021166-print-pdf.pdf_