In 2017, Bill Gates posed a simple but powerful question: If a factory worker who produces $50,000 worth of output a year pays income and social security taxes, why shouldn’t a robot that takes over the same job bear a similar tax burden? Gates argued that a robot tax could both help manage the pace of automation and finance the retraining of workers who lose their jobs. 1 From an economic perspective, however, the issue is more fundamental than whether robots should pay taxes: Why do we tax labor and capital differently?
The standard optimal taxation approach holds that taxes should distort production decisions as little as possible.2 3 A robot, after all, is a capital good. Taxing it simply because it replaces workers could artificially raise the cost of using robots and influence firms’ investment and technology choices. At the same time, if the tax burden changes when a firm substitutes capital for labor in the same production activity, then the tax system may already be influencing its choice of technology.4 The fact that automation may worsen income inequality does not necessarily imply that robots should be taxed separately.5 Rather than discouraging the use of productive technologies through taxation, the standard approach favors addressing the distributional consequences of automation through the broader tax-and-transfer system.
But what if the existing tax-and-transfer system is not neutral to begin with?
Calculations by Acemoglu, Manera, and Restrepo for the United States point precisely to this problem. While the average tax burden on labor has been around 25 percent, the effective tax burden on capital goods such as software and equipment has fallen substantially over time. A similar concern arises in Türkiye: while taxes and social security contributions impose a significant burden on labor, certain investments in machinery benefit from depreciation allowances and investment incentives.6 7 8In some cases, this can make the use of capital more attractive than labor from a tax perspective. As a result, the tax system may encourage firms to “use machines instead of workers” for certain tasks even when doing so generates little or no productivity gain. In other words, technology may replace labor not only because it is more efficient, but also because the tax system makes it relatively cheaper.
At this point, a robot tax begins to look more reasonable. Yet the optimal taxation literature remains cautious. According to Thuemmel, a robot tax can reduce income inequality under certain conditions. But this does not necessarily mean that robots need a special tax. Similar distributional goals may be achieved more effectively, and with less economic cost, by improving the income tax system.Moreover, taxing robots is not always the optimal policy. Thuemmel shows that the appropriate tax depends on how robots affect different workers’ wages. If automation mainly puts downward pressure on the wages of lower-paid workers, taxing robots can help reduce inequality. But if robots raise the relative wages of lower-paid workers, the logic can reverse: subsidizing robots may instead improve income distribution. In other words, whether robots should be taxed or subsidized depends on who gains and who loses from their adoption.9
This is where the fundamental trade-off of a robot tax becomes clear. IMF estimates suggest that taxing robots can slow the pace at which companies replace workers with machines and thereby reduce income inequality.10 But making robots more expensive can also weaken firms’ incentives to invest in new technologies. The long-run cost may be lower investment, slower productivity growth, and less output.11
So where does this leave us?
We should first ask why the tax system may make employing people more expensive than using machines. The answer may not be a new tax on robots. Instead, governments could reduce the tax burden on workers, tax capital income and excess profits more effectively, and support workers who lose income because of automation. This would make it possible to address the distributional consequences of automation without directly raising the cost of using robots and potentially slowing investment and technological progress. Indeed, in the age of artificial intelligence, the IMF recommends strengthening the taxation of capital income, capital gains, and excess profits rather than introducing a new tax specifically targeting technology.12
The fact that robots do not pay taxes may not be the real problem. The real problem may be that, in some cases, the tax system makes using robots more attractive than employing people.
REFERENCES
- Delaney, K. J. (2017, February 20). Bill Gates: This is why we should tax robots. World Economic Forum. https://www.weforum.org/stories/emerging-technologies/bill-gates-this-is-why-we-should-tax-robots/ ↩︎
- Gordon, R. (2011). Commentary on Tax by Design: The Mirrlees Review. Fiscal Studies, 32(3), 395–414. https://ifs.org.uk/journals/commentary-tax-design-mirrlees-review ↩︎
- Keuschnigg, C. (2011). The design of capital income taxation: Reflections on the Mirrlees Review. Fiscal Studies, 32(3), 437–452. https://ifs.org.uk/journals/design-capital-income-taxation-reflections-mirrlees-review ↩︎
- Organisation for Economic Co-operation and Development. (2018). Addressing labour-market disruptions from trade and automation. In OECD economic surveys: United States 2018. OECD Publishing. https://www.oecd.org/en/publications/oecd-economic-surveys-united-states-2018_eco_surveys-usa-2018-en/full-report/component-5.html ↩︎
- Acemoglu, D., Manera, A., & Restrepo, P. (2020). Does the U.S. tax code favor automation? (NBER Working Paper No. 27052). National Bureau of Economic Research. https://www.nber.org/papers/w27052 ↩︎
- Organisation for Economic Co-operation and Development. (2026). Taxing Wages 2026: Türkiye. OECD Publishing. https://www.oecd.org/content/dam/oecd/en/publications/reports/2026/04/taxing-wages-2026-country-notes_491a0e97/turkiye_522f162c/e91bacbe-en.pdf
↩︎ - Republic of Türkiye, Ministry of Industry and Technology. (2025). Yatırımlarda devlet yardımları hakkında karar [Decision on state aid for investments] (Decision No. 9903). https://sanayi.gov.tr/assets/pdf/mevzuat/YatirimlardaDevletYardimlariHakkindaKarar9903.pdf ↩︎
- Republic of Türkiye, Revenue Administration. (n.d.). Vergi Usul Kanunu [Tax Procedure Law] (Law No. 213). İstanbul Defterdarlığı.
https://istanbul.gib.gov.tr/mevzuat/kanun/434 ↩︎ - Thuemmel, U. (2023). Optimal taxation of robots. Journal of the European Economic Association, 21(3), 1154–1190. https://academic.oup.com/jeea/article/21/3/1154/6798383 ↩︎
- Berg, A., Bounader, L., Gueorguiev, N., Miyamoto, H., Moriyama, K., Nakatani, R., & Zanna, L.-F. (2021). For the benefit of all: Fiscal policies and equity-efficiency trade-offs in the age of automation (IMF Working Paper No. 2021/187). International Monetary Fund. https://www.elibrary.imf.org/view/journals/001/2021/187/article-A001-en.xml ↩︎
- Gueorguiev, N., & Nakatani, R. (2021, November 18). Sharing the gains of automation: The role of fiscal policy. IMF Blog. https://www.imf.org/en/blogs/articles/2021/11/18/sharing-the-gains-of-automation-the-role-of-fiscal-policy ↩︎
- Brollo, F., Dabla-Norris, E., de Mooij, R., Garcia-Macia, D., Hanappi, T., Liu, L., & Nguyen, A. D. M. (2024). Broadening the gains from generative AI: The role of fiscal policies (IMF Staff Discussion Note No. 2024/002). International Monetary Fund. https://www.elibrary.imf.org/view/journals/006/2024/002/article-A001-en.xml ↩︎
Leave a Reply