## To reverse widening inequality, keep a tight rein on automation

## Source details

**Canonical URL:** [To reverse widening inequality, keep a tight rein on automation](https://www.imf.org/-/media/files/publications/fandd/article/2021/march/en/covid-inequality-and-automation-acemoglu-en.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/fandd/article/2021/march/en/covid-inequality-and-automation-acemoglu-en.pdf.md)
- [Structured JSON version](/-/media/files/publications/fandd/article/2021/march/en/covid-inequality-and-automation-acemoglu-en.pdf.json)

---

### Diagnosis: automation and unshared growth
- Economic growth since the 1980s has become "much less shared": wider inequality, disappearance of good high-paying secure jobs, and decline in the real wages of less-educated workers in the United States.
- Research with Pascual Restrepo indicates automation accounts for much of this loss of shared growth, alongside globalization and the declining power of labor relative to capital.
- The next phase of automation, driven by machine learning and artificial intelligence (AI), could either further exacerbate inequality or—if directed by government policy—contribute to a resumption of shared growth.

### Mechanisms driving excessive automation
- Automation defined: substitution of machines and algorithms for tasks previously performed by labor.
- Employers do not internalize the social disruption caused by job losses, creating a bias toward excessive automation.
- Corporate strategy concentration:
  - A handful of very large tech companies have tiny workforces and business models built on automation.
  - Big Tech companies including Amazon, Alibaba, Alphabet, Facebook, and Netflix are responsible for more than $2 of every $3 spent globally on AI (McKinsey Global Institute 2017).
  - Their vision influences other firms, researchers, and students, skewing technological effort toward substitution of algorithms for humans.
- Tax policy incentives:
  - Over the past 40 years, labor has paid an effective tax rate of more than 25 percent via payroll and federal income taxes (Acemoglu, Manera, and Restrepo 2020).
  - About 20 years ago, equipment and software investment faced tax rates of about 15 percent.
  - Today, investments in software and equipment are taxed at rates of less than 5 percent in many cases, and corporations can even derive net subsidies when they invest in capital—creating a powerful motive for excessive automation.

### Consequences for productivity and labor
- Despite rapid advances in information and communication technology, total factor productivity growth in the US has been very low and much lower over the past 20 years than during the decades after World War II (Gordon 2017).
- Many automation technologies (self-checkout kiosks, automated customer service) are not generating much total factor productivity growth.
- Excessive automation outcomes:
  - Adoption beyond what reduces production costs.
  - Social costs via lower employment and worker wages.
  - Potential contribution to the slowdown in productivity growth by diverting attention away from technologies and organizational changes that are complementary to human labor.

### Policy prescriptions to redirect technological change
- Objective: tilt the composition of innovation away from excessive automation and toward human-friendly technologies that produce employment opportunities and more shared prosperity.
- Policy levers:
  - Correct asymmetric taxation that favors capital over labor (first step: address the differential tax treatment that encourages automation).
  - Use R&D subsidies targeted to technologies that help human productivity and increase labor demand.
  - Government purchases and modest subsidies can redirect technological trajectories, as illustrated by renewable energy development.
- Sectoral opportunities for human-friendly technologies:
  - Education: AI for adaptive, student-centered teaching combined with better-trained teachers.
  - Health care: AI and digital technologies to empower nurses and technicians to provide more and better services.
  - Modern manufacturing: augmented reality and computer vision to increase human productivity in production.
  - Digital communications: technologies like Zoom broaden human communication and capabilities.

### Political and institutional considerations
- Governments already influence technological direction through tax policies and support for corporate research and universities; redirection is not unprecedented.
- Historical example: renewable energy shift
  - Renewables make up 19 percent of energy consumption in Europe and 11 percent in the United States.
  - Cost declines brought renewables into cost parity with fossil-fuel energy through targeted subsidies and changing consumer norms (IRENA 2019; IRENA 2020).
- Challenges and safeguards:
  - Concerns about excessive state direction (Hayekian critique) countered by historical experience: welfare states did not crush liberty but often deepened democracy.
  - Successful redirection requires strengthening society’s capacity to hold the state accountable—mobilization of citizens, active civil society, informed media.
  - New digital technologies can weaken democratic checks (misinformation, AI-powered social media creating filter bubbles), making societal safeguards more urgent.
- Additional policy objective: address dominance of a handful of big tech companies to ensure greater competition and protect privacy while influencing the direction of future technology.

*Source: Daron Acemoğlu, "To reverse widening inequality, keep a tight rein on automation," Finance & Development, March 2021.*

---


_Source: https://www.imf.org/-/media/files/publications/fandd/article/2021/march/en/covid-inequality-and-automation-acemoglu-en.pdf_