Technology and Social Change

Lecture: 11 — Technology, Labor & Automation

Bogdan G. Popescu
bgpopescu@tec.mx

Tecnológico de Monterrey

Lecture Overview

  • The “automation anxiety” puzzle
  • Task-based models of technological change
  • Displacement, reinstatement, and distribution
  • Institutional mediation of tech impact
  • Comparative and historical evidence
  • AI, platforms, and policy responses

Three Motivating Puzzles

Puzzle 1 — Productivity vs. Employment

If automation kills jobs, why has employment risen for two centuries?

Puzzle 2 — Growth vs. Inequality

Why does rising output often increase inequality?

Puzzle 3 — The Hollowing Middle

Why do middle-skill jobs shrink while extremes grow?

Learning Objectives

  1. Explain the task-based framework (Autor, Acemoglu)
  2. Distinguish automation from augmentation
  3. Identify causal mechanisms linking tech to labor
  4. Analyze how institutions mediate inequality
  5. Evaluate determinism vs. contingency
  6. Apply frameworks to AI, platforms, and policy

Conceptual Framework

Why Simple Automation Models Fail

The “Luddite fallacy” predicts outcomes we do not observe:

  • Fixed demand assumption — ignores price-driven demand growth
  • Ignores new task creation — tech creates occupations too
  • Treats labor as homogeneous — workers differ in tasks
  • Neglects institutions — same tech, different outcomes

“Confusing task displacement with employment effects.” — Acemoglu & Restrepo (2019)

So What?

If simple models fail, we need a better framework.

Next: The task-based approach decomposes jobs into their component tasks — revealing how technology actually reshapes work.

The Task-Based Framework

Key Insight

Jobs are bundles of tasks. Technology affects tasks, not jobs directly.

┌────────────────────────────────────────────────────────┐
│                    JOB DECOMPOSITION                   │
│                                                        │
│    "Accountant" = { data entry, analysis, client       │
│                     communication, judgment calls }    │
│                                                        │
│    Automation affects these tasks differentially:      │
│    • Data entry: HIGH substitution potential           │
│    • Analysis: PARTIAL augmentation                    │
│    • Communication: LOW substitution potential         │
│    • Judgment: VERY LOW substitution potential         │
└────────────────────────────────────────────────────────┘

Autor’s Task Taxonomy

Task Type Examples Automation Risk
Routine Cognitive Data entry, bookkeeping HIGH
Routine Manual Assembly, machine operation HIGH
Non-Routine Analytical Research, programming VARIABLE
Non-Routine Interpersonal Teaching, negotiation LOW
Non-Routine Manual Janitorial, food service LOW

Source: Autor, Levy & Murnane (2003)

The Acemoglu-Restrepo Framework

Technology produces dual effects on labor:

  1. Displacement — capital replaces labor in existing tasks
  2. Reinstatement — new tasks where labor has advantage
  3. Productivity — output expands from efficiency gains

The net employment effect is empirically indeterminate — it depends on which force dominates.

Displacement-Reinstatement Model

%%{init: {'theme': 'base', 'themeVariables': {'fontSize': '20px', 'primaryColor': '#4a7c6f', 'primaryTextColor': '#1e293b', 'primaryBorderColor': '#334155', 'lineColor': '#64748b', 'secondaryColor': '#f9fafb', 'tertiaryColor': '#f9fafb'}, 'flowchart': {'useMaxWidth': true, 'nodeSpacing': 40, 'rankSpacing': 50, 'width': 1150, 'height': 650}}}%%
flowchart LR
    A["Technological<br/>Change"] --> B["Displacement<br/>Effect"]
    A --> C["Reinstatement<br/>Effect"]
    A --> D["Productivity<br/>Effect"]
    B --> E["Automates existing tasks<br/>(− Labor Demand)"]
    C --> F["Creates new tasks<br/>(+ Labor Demand)"]
    D --> G["Expands output<br/>(± Labor Demand)"]
    E --> H{"Net Employment<br/>Effect"}
    F --> H
    G --> H
    style A fill:#1e293b,color:#f9fafb,stroke:#334155
    style B fill:#b44527,color:#f9fafb,stroke:#334155
    style C fill:#4a7c6f,color:#f9fafb,stroke:#334155
    style D fill:#b7943a,color:#1e293b,stroke:#334155
    style H fill:#64748b,color:#f9fafb,stroke:#334155

Source: Based on Acemoglu & Restrepo (2019)

Automation vs. Augmentation

Dimension Automation Augmentation
Definition Machines replace human labor Machines complement human labor
Labor effect Substitution (demand ↓) Complementarity (demand ↑)
Skill bias Reduces routine-skill demand Raises cognitive-skill demand
Examples ATMs, assembly robots CAD software, diagnostic AI

Source: Based on Acemoglu (2021)

“Same technology can be automation or augmentation — a design choice.” — Acemoglu (2021)

Why Excessive Automation?

The tax system creates bias toward automation:

  • Labor is heavily taxed (income, payroll, benefits)
  • Capital faces lower effective tax rates
  • Firms substitute capital even when socially inefficient
  • Result: “so-so automation” — displaces without big gains

So What?

The framework shows that automation is not destiny — it depends on which effect dominates and on design choices.

Next: What specific causal mechanisms link technology to inequality and labor market polarization?

Exercise 1: Task Decomposition

Prompt: Pick a job you know well (barista, teacher, journalist, nurse). Decompose it into 4–5 tasks. For each task, rate automation potential as HIGH / MEDIUM / LOW. Which tasks might AI affect that traditional automation could not?

Estimated time: 5 minutes. Discuss in pairs, then share.

Causal Mechanisms

The Causal System

%%{init: {'theme': 'base', 'themeVariables': {'fontSize': '20px', 'primaryColor': '#4a7c6f', 'primaryTextColor': '#1e293b', 'primaryBorderColor': '#334155', 'lineColor': '#64748b', 'secondaryColor': '#f9fafb', 'tertiaryColor': '#f9fafb'}, 'flowchart': {'useMaxWidth': true, 'nodeSpacing': 50, 'rankSpacing': 50, 'width': 1150, 'height': 650}}}%%
flowchart LR
    A["Institutions<br/>(taxes, unions,<br/>education)"] -->|shape| B["Technological<br/>Change"]
    B --> C["Task<br/>Reconfiguration"]
    C --> D["Labor Market<br/>Adjustment"]
    D --> E["Distributional<br/>Outcomes"]
    E --> F["Political<br/>Responses<br/>(policy, parties)"]
    F -->|reshape| A

    style A fill:#4a7c6f,color:#f9fafb,stroke:#334155
    style B fill:#1e293b,color:#f9fafb,stroke:#334155
    style F fill:#b44527,color:#f9fafb,stroke:#334155

Causation is recursive: institutions shape technology and are reshaped by its consequences.

Mechanism 1: Skill-Biased Technological Change

Definition

SBTC: Technologies that raise the relative productivity and wages of skilled workers compared to unskilled workers.

Causal pathway: Computer tech → ↑ cognitive-skill demand → ↑ college premium → ↑ inequality

U.S. College Wage Premium, 1963–2022

Figure 1: College vs. high-school earnings gap

Mechanism 2: Routine-Biased Technological Change

SBTC cannot explain job polarization. Autor (2003, 2013) proposes RBTC:

  • Computers excel at rule-based tasks
  • Routine tasks (cognitive AND manual) are automatable
  • Non-routine tasks at BOTH ends are complemented
  • Middle-skill, routine jobs decline
  • High-skill AND low-skill service jobs grow

U.S. Job Polarization, 1980–2005

Figure 2: Employment share change by skill percentile

Mechanism 3: Geographic Concentration

Automation effects are spatially concentrated:

  • Manufacturing automation devastated specific regions
  • Knowledge economy concentrates in “superstar cities”
  • Local labor markets adjust slowly
  • Geographic immobility compounds individual displacement

Import competition from China destroyed approximately 2.4 million jobs across the broader economy, with effects concentrated in specific commuting zones (Autor, Dorn & Hanson, 2013, 2016).

Mechanism 4: Productivity-Wage Decoupling

Figure 3: Productivity grew far faster than compensation

Why Do Wages Lag Productivity?

  1. Declining labor share — capital captures more output
  2. Within-labor inequality — top earners capture gains
  3. Monopsony power — employer concentration suppresses wages
  4. Institutional erosion — weaker unions, stagnant minimums
  5. Globalization — trade compounds tech displacement

So What?

Four mechanisms — SBTC, RBTC, geography, and decoupling — explain how technology reshapes labor markets. But outcomes vary enormously across countries.

Next: Institutions determine whether the same technology produces moderate or extreme inequality.

Institutions & Path Dependence

The Institutional Mediation Thesis

Key Claim

The same technology produces different outcomes across institutional contexts. Effects on employment, wages, and inequality are institutionally contingent.

Cross-National Evidence

Figure 4: High robot density does not guarantee high inequality

Same Technology, Different Outcomes

%%{init: {'theme': 'base', 'themeVariables': {'fontSize': '18px', 'primaryColor': '#4a7c6f', 'primaryTextColor': '#1e293b', 'primaryBorderColor': '#334155', 'lineColor': '#64748b', 'secondaryColor': '#f9fafb', 'tertiaryColor': '#f9fafb'}, 'flowchart': {'useMaxWidth': true, 'nodeSpacing': 30, 'rankSpacing': 45, 'width': 1150, 'height': 650}}}%%
flowchart TD
    A["Industrial Robots<br/>(Same Technology)"] --> B["Germany"]
    A --> C["United States"]
    A --> D["Sweden"]
    B --> E["Strong unions<br/>Vocational ed<br/>Codetermination"]
    C --> F["Weak unions<br/>Limited training<br/>Market-led"]
    D --> G["Strong unions<br/>Active labor<br/>market policy"]
    E --> H["Moderate inequality<br/>Job security"]
    F --> I["Sharp inequality<br/>Job loss"]
    G --> J["Low inequality<br/>Smooth transitions"]
    style A fill:#1e293b,color:#f9fafb,stroke:#334155
    style H fill:#b7943a,color:#1e293b,stroke:#334155
    style I fill:#b44527,color:#f9fafb,stroke:#334155
    style J fill:#4a7c6f,color:#f9fafb,stroke:#334155

Source: Conceptual diagram based on Hall & Soskice (2001); Thelen (2014)

Path Dependence and Lock-In

Brian Arthur’s Path Dependence

Early institutional choices become self-reinforcing, making later changes difficult even when alternatives are superior.

U.S. path: Limited union institutionalization (1930s–50s) → weak safety net → few policy tools for managing automation today

Germany path: Codetermination established (1940s–60s) → strong works councils → technology adoption shaped by worker input

Source: Arthur (1989); Pierson (2004)

Mokyr: Technology and Institutional Adaptation

Mokyr (1990, 2002) emphasizes:

  • Technology alone does not determine outcomes
  • “Useful knowledge” requires institutional support
  • Industrial Revolution needed complementary institutions
  • Patent systems, education, and labor regulation co-evolved
  • Implication: today’s transition requires institutional innovation

So What?

Institutions — unions, education, tax policy — explain why Germany and Sweden manage automation better than the U.S. despite similar robot density.

Next: What does the empirical evidence actually show about robots, trade, and employment?

Exercise 2: Institutional Design

Prompt: You are advising Mexico’s government on automation policy. Mexico has moderate union density and a large informal sector. Drawing on the Germany/Sweden/U.S. comparison, which 2–3 institutional features would you prioritize and why?

Estimated time: 5 minutes. Small-group discussion.

Empirical Evidence

Acemoglu & Restrepo (2020): Robots and Jobs

Research design:

  • Variation in robot adoption across U.S. commuting zones
  • Instrument with European industry robot adoption
  • Estimate employment and wage effects
Outcome Effect per Robot / 1,000 Workers
Employment-to-population ratio −0.2 pp
Wages −0.42%
Manufacturing jobs Disproportionate decline

Source: Acemoglu & Restrepo (2020)

Robot Adoption and Employment

Figure 5: Higher robot penetration associated with employment decline

The China Shock

Autor, Dorn & Hanson (2013, 2016, 2021):

  • Import competition caused ~2.4 million U.S. job losses
  • Effects concentrated in specific local labor markets
  • Workers did not relocate; many left labor force
  • Wages declined in exposed regions
  • Effects persisted for over a decade

Political consequence: Trade-exposed regions shifted toward populist candidates (Autor et al., 2020).

Frey & Osborne (2017): Automation Risk

Headline finding:

“47% of U.S. employment at high risk of automation.”

Critiques:

  • Estimates at occupation-level, not task-level
  • Ignores economic and institutional barriers
  • OECD re-analysis (task-based): only 9% at high risk
  • Arntz, Gregory & Zierahn (2016): similar revision

Empirical Consensus

What the Evidence Actually Shows

  1. Automation does reduce employment in affected tasks/places
  2. Effects are localized and persistent, not rapidly dispersed
  3. Aggregate employment has not collapsed — reinstatement is real
  4. Inequality rose substantially; tech is a contributing factor
  5. Institutional context shapes the magnitude dramatically
  6. Adjustment is slow and costly for displaced workers

So What?

The evidence confirms both displacement and reinstatement are real. But adjustment is slow, localized, and painful — and institutions matter enormously.

Next: How should we interpret all this? Is technology destiny, or is it shaped by politics?

Competing Interpretations

Interpretation 1: Technological Determinism

Technology has autonomous logic driving social outcomes independent of human choice.

  • Exponential progress creates inexorable pressure
  • AI represents a qualitative break with past automation
  • Cognitive tasks now automatable (“this time is different”)
  • Resistance is futile; adaptation is the only option

Proponents: Brynjolfsson & McAfee (qualified); Kurzweil (strong form)

Interpretation 2: Institutional Contingency

Technology creates possibilities; institutions determine outcomes.

  • Same technologies → different outcomes across societies
  • Technology adoption is itself an institutional choice
  • Policy can redirect technological development
  • Historical transitions were managed through institutional innovation

Proponents: Acemoglu, Autor, Mokyr, Piketty, Hall & Soskice

The Debate Matrix

Dimension Determinism Contingency
Causation Tech → Outcomes Tech ↔︎ Institutions
Agency Limited Substantial
Policy Adapt to the inevitable Shape the development path
AI Qualitative break Continuous with history
Prediction Massive displacement Depends on choices

Source: Author’s synthesis

An Integrated View

Synthesis

Technology creates pressures and possibilities, but institutions determine the rate, direction, distribution, and adjustment mechanisms.

The question is not “Will AI take our jobs?” but:

“What institutions will we build so progress benefits broadly?”

So What?

Neither pure determinism nor pure choice captures reality. Technology constrains; institutions choose within those constraints.

Next: How do these frameworks apply to today’s challenges — AI, platforms, and policy?

Modern Relevance

AI and Large Language Models

How does AI fit the task-based framework?

  • AI automates cognitive, non-routine tasks (new frontier)
  • RBTC extension: writing, analysis, coding now affected
  • Strong complementarity potential for knowledge workers
  • Speed of change faster than previous transitions
  • Novel capabilities make prediction difficult

Platform Labor and the Gig Economy

Platform Dynamics

Digital platforms restructure employment relationships in ways that often shift risk to workers.

  • Independent contractor status avoids labor protections
  • Algorithmic management replaces human supervisors
  • Task-level labor markets atomize work
  • Surveillance and control without employment rights

Examples: Uber, DoorDash, Mechanical Turk, Upwork

Policy Responses: A Typology

Approach Examples Logic
Laissez-faire Minimal intervention Markets adjust
Social insurance Retraining, UI benefits Cushion transitions
UBI Universal basic income Decouple income from work
Taxation Robot taxes, wealth taxes Slow automation; redistribute
Tech policy Direct R&D toward augmentation Change the technology path
Labor institutions Strengthen unions, codetermination Rebalance power

Source: Synthesized from Acemoglu (2021); Korinek & Stiglitz (2021)

The Political Economy of Automation

  • Automation exposure predicts populist voting (Autor et al., 2020)
  • Robot exposure correlates with Trump support (Frey et al., 2018)
  • Rising inequality fuels political instability (Piketty, 2020)
  • Mechanism: displacement → status anxiety → anti-establishment politics

Rodrik (2018) argues that populism is, in part, a political response to the uneven distributional consequences of globalization and technological change.

So What?

AI extends the automation frontier to cognitive work. Without institutional adaptation, the political backlash we have already seen will intensify.

Next: What are the key takeaways and how do we synthesize everything?

Exercise 3: AI Policy Debate

Prompt: Should governments impose a tax on AI systems that replace human workers? Split into two groups: one argues FOR (citing Acemoglu’s “so-so automation”), the other AGAINST (citing innovation effects). Each group prepares 3 arguments.

Estimated time: 5 minutes. Structured debate format.

Synthesis & Conclusion

Key Takeaways (1/2)

  1. Jobs = bundles of tasks — tech affects tasks differentially
  2. Dual effects: displacement (−) and reinstatement (+) are both real
  3. RBTC explains polarization — routine tasks automated regardless of skill
  4. Institutions mediate outcomes — same tech, different inequality

Key Takeaways (2/2)

  1. Adjustment is slow and costly — markets do not rapidly equilibrate
  2. Path dependence constrains policy — history shapes current options
  3. AI extends the frontier — cognitive tasks now affected
  4. Policy is choice — automation effects are institutionally shaped

Discussion Questions

  1. Which mechanism (SBTC, RBTC, geography, institutions) best explains contemporary inequality? Why?

  2. Why might Germany experience lower inequality from automation than the U.S.?

  3. Is AI qualitatively different from previous automation, or a continuation?

  4. How should policymakers evaluate the trade-off between a robot tax and innovation?

  5. If automation concentrates gains among owners, what principles should guide redistribution?

Readings

References

  • Acemoglu, D. (2021). Harms of AI. NBER Working Paper No. 29247. National Bureau of Economic Research.

  • Acemoglu, D., & Restrepo, P. (2019). Automation and new tasks: How technology displaces and reinstates labor. Journal of Economic Perspectives, 33(2), 3–30.

  • Acemoglu, D., & Restrepo, P. (2020). Robots and jobs: Evidence from U.S. labor markets. Journal of Political Economy, 128(6), 2188–2244.

  • Arntz, M., Gregory, T., & Zierahn, U. (2016). The risk of automation for jobs in OECD countries: A comparative analysis. OECD Social, Employment and Migration Working Papers, No. 189. OECD Publishing.

  • Arthur, W. B. (1989). Competing technologies, increasing returns, and lock-in by historical events. Economic Journal, 99(394), 116–131.

  • Autor, D. H., & Dorn, D. (2013). The growth of low-skill service jobs and the polarization of the U.S. labor market. American Economic Review, 103(5), 1553–1597.

  • Autor, D. H., Dorn, D., & Hanson, G. H. (2013). The China syndrome: Local labor market effects of import competition in the United States. American Economic Review, 103(6), 2121–2168.

  • Autor, D. H., Dorn, D., & Hanson, G. H. (2016). The China shock: Learning from labor-market adjustment to large changes in trade. Annual Review of Economics, 8, 205–240.

References

  • Autor, D. H., Dorn, D., Hanson, G. H., & Majlesi, K. (2020). Importing political polarization? The electoral consequences of rising trade exposure. American Economic Review, 110(10), 3139–3183.

  • Brynjolfsson, E., & McAfee, A. (2014). The second machine age: Work, progress, and prosperity in a time of brilliant technologies. W. W. Norton.

  • Frey, C. B., Berger, T., & Chen, C. (2018). Political machinery: Did robots swing the 2016 U.S. presidential election? Oxford Review of Economic Policy, 34(3), 418–442.

  • Frey, C. B., & Osborne, M. A. (2017). The future of employment: How susceptible are jobs to computerisation? Technological Forecasting and Social Change, 114, 254–280.

  • Goldin, C., & Katz, L. F. (2008). The race between education and technology. Harvard University Press.

  • Hall, P. A., & Soskice, D. (2001). Varieties of capitalism: The institutional foundations of comparative advantage. Oxford University Press.

  • Korinek, A., & Stiglitz, J. E. (2021). Artificial intelligence, globalization, and strategies for economic development. NBER Working Paper No. 28453. National Bureau of Economic Research.

  • Kurzweil, R. (2005). The singularity is near: When humans transcend biology. Viking.

  • Mokyr, J. (1990). The lever of riches: Technological creativity and economic progress. Oxford University Press.

References

  • Mokyr, J. (2002). The gifts of Athena: Historical origins of the knowledge economy. Princeton University Press.

  • Pierson, P. (2004). Politics in time: History, institutions, and social analysis. Princeton University Press.

  • Piketty, T. (2020). Capital and ideology. Harvard University Press.

  • Rodrik, D. (2018). Populism and the economics of globalization. Journal of International Business Policy, 1(1–2), 12–33.

  • Thelen, K. (2014). Varieties of liberalization and the new politics of social solidarity. Cambridge University Press.

  • Autor, D. H., Levy, F., & Murnane, R. J. (2003). The skill content of recent technological change: An empirical exploration. Quarterly Journal of Economics, 118(4), 1279–1333.