The rise of robots, employment, income inequality, and optimal taxation, GFP Ottawa 2018

1. THE RISE OF ROBOTS,
EMPLOYMENT, INCOME
INEQUALITY, AND OPTIMAL
TAXATION

2. • Automation has the potential to make obsolete a large proportion
of workers (14% of jobs highly automatable, 32% face substantial
change – source: OECD)
• Routine workers (e.g. low skills) are most a risk of losing their
jobs and of lower wages.
• Non-Routine jobs (software engineers, researchers, innovators)
are less at risk and benefit from rising skill premiums.
• Hence, concerns that automation wave could contribute to higher
non-employment and/or income inequality.
• With frictions, incomplete markets, and imperfect information, the
outcome may not be efficient.
• Loss of tax revenues from labour prevents sharing the benefits
from winners to losers.
• Already visible regional concentration of joblessness and low
participation (Glaeser, Summers and Austin, 2018).
Diffusion of robots and automation has
given rise to debates

3. Key business leaders and scientists have
called to tax the robots (e.g. Gates, Musk,
Hawking). Objectives:
• Distort prices to protect routine workers
• Transfer revenue to vulnerable workers
• Pay for universal basic income
=> Example of South Korea “robot tax”
(lower tax deduction from investment in
automation)
A tax policy response?

4. • In context of OECD Economic Survey of
the USA, work was commissioned to
Julien Daubanes (Copenhagen) and
Pierre-Yves Yanni (Montreal)
• Project uses the framework of Guerreiro,
Rebelo & Teles (2018)
• Theoretical framework + simulation of
comparative tax policies
The paper

5. • Intermediate goods used in the production of
other goods, such as robots, should not be taxed
because this reduces the efficiency of
production; better to tax consumption or income
• However, the theorem relies on disputed
assumptions.
Economist’s views of optimal taxation

6. • Even when the government needs to
implement distortionary taxes, it should not
distort production.
• Taxing consumption and income at different
rates is more optimal.
• This requires incentive-compatible
mechanisms with workers truthfully
reporting their type of labour.
• In practice, governments not aware of labour
types => non-routine workers could behave
like routine workers.
Diamond & Mirrlees, 1971

7. • Indirect taxes cannot redistribute more
efficiently than the optimal income tax
system.
• Critical assumption: Workers with
different productivities are perfect
substitutes in production.
• In reality, routine workers and non-
routine workers are not substitutable.
Atkinson & Stiglitz 1976

8. • X is quantity of robots
• Nr is quantity of routine labour,
substitutable to X
• Nn is quantity of non-routine labour,
complement of Nr and Nn
• Production function
• Households max utility
MODEL

9. When the cost of robots decrease:
• The number of robots increase
• Production increase
• Labor income of routine workers is reduced
• This is more than compensated by increase in non-
routine labor income
• Utility levels benefit from higher consumption (non-
routine workers) and reduced working hours (routine
workers).
• Non-routine workers contribute more to tax revenue
• Inequality increases.
Rise of robots with progressive income
taxes: key results

10. Comparison of steady states

11. • Automation increases from partial to complete
• Increasing wage inequality is compensated by increasing
redistributive transfers
• Utility increases but is lower for non-routine workers (work
more)
First-best tax policy:
Robots are not taxed, transfers n -> r

12. Difficult to make type-specific transfers
• Governments observe income, not type of
labour
• Non-routine worker can obtain the same
consumption by working less if it mimics
the routine worker.
• Hence, incentive compatibility constraint
is not satisfied
Practical difficulties with first-best policy

13. • Taxing robots alleviates the incentive
compatibility constraint by encouraging non-
routine worker to behave as such.
• Thus, this makes the income-transfer system
more efficient.
• Useful only when automation is incomplete,
i.e. when there are routine workers. If full
automation=> not tax on robots is needed,
but inequality needs to be addressed.
Second-best tax policy: taxing robots

14. After-tax income is:
• λ controls the level of taxation and γ the
progressivity
• Ω is a lump-sum transfer, or universal
income
Second-best with universal income

15. With universal income transfer

16. As cost of robots decreases
• Total revenues increase, hence the level of income taxation
should decrease
• Income taxation should become more progressive, in the face
of rising inequality
• A universal income could be implemented
• However, many difficulties with universal income schemes:
happiness does not come only from material wellbeing – work
also provides sense of accomplishment (Glaeser, 2018).
• Can society function with 50% of adults who are not
participating in labour market and live on social benefits?
Pros and cons of
universal basic income

17. When routine workers are transformed into
non-routine workers
• There are more robots
• Nobody is worse-off
• Production and revenue from taxes
increases
• But inequality increases
Extension with training of routine
workers

18. With rise of robots
• Tax revenues can be expected to increase
• Inequality will increase
• Universal income and robot taxes more
efficient than higher income taxes
In practice, many improvements to tax systems
can be done before these options are
considered
• Tax economic rents from natural resources
• Improve transfer system (i.e. more targeted)
Key takeaways