← Articles
USA version
USAUKAustralia

The 10% Problem: Why AI Job Displacement Will Break National Budgets Faster Than Anyone Expects

By Daniel Horan


Economists and governments are engaged in a furious debate about how many jobs artificial intelligence will destroy. They are asking the wrong question.

The number that should be keeping finance ministers awake at night is not the percentage of jobs that AI will eliminate. It is what happens to a national budget when it does. Because when you trace the fiscal mechanics of even a modest displacement — say, 10% of the formal workforce — what you find is not a 10% problem. You find a system failure.

The mathematics are brutal, and largely absent from public debate.


A Government Is a Fixed-Cost Machine

Start with an uncomfortable structural reality. Modern governments — whether in Washington, Manila, or Berlin — are not flexible organisations. They are fixed-cost machines. In the United States, roughly 60% of all federal spending is legally mandated: Social Security, Medicare, Medicaid, and debt interest. It runs on autopilot, set by formula, immune to annual budget decisions. Add net interest payments and you are looking at nearly 75 cents of every government dollar that cannot be quickly reduced regardless of what happens to revenue.

This is not a uniquely American condition. Across OECD nations, mandatory entitlement spending has grown steadily as a share of GDP for four decades, while discretionary spending has shrunk. Governments have quietly traded flexibility for obligation.

When revenue falls, a government cannot simply scale back like a business cutting headcount. It must borrow to meet its fixed commitments — and that borrowing compounds the problem.


Where Government Revenue Actually Comes From

Now consider where that revenue originates.

In most developed economies, labour — wages, salaries, and the payroll taxes attached to them — accounts for approximately 75% of government tax receipts. In the United States, individual income taxes and payroll taxes together represent roughly 85% of federal revenue. The picture is similar across Europe. Corporate taxes, capital gains, and consumption taxes make up the remainder.

This is the structural exposure that AI creates. It is not simply that displaced workers stop paying income tax. It is that the entire fiscal architecture of the modern state was built on the assumption of mass formal employment — and that assumption is now being quietly dismantled.


The Cascade: Why 10% Becomes 25%

Here is what a 10% workforce displacement actually produces fiscally, step by step.

Step one — direct revenue loss. Ten percent of the workforce no longer earning wages means approximately 7.5% of total government revenue disappears immediately, given labour's three-quarters share of the tax base. But this is the smallest part of the problem.

Step two — the consumption collapse. Displaced workers stop spending. Consumer spending drives VAT receipts, sales taxes, and the corporate profits — and thus corporate taxes — of the businesses that serve them. A Brookings Institution analysis found that even modest labour displacement "could significantly strain public finances" through precisely this consumption channel. The revenue loss compounds beyond the initial wage-tax reduction.

Step three — mandatory spending surges. At the exact moment revenue is falling, government expenditure must rise. Displaced workers claim unemployment benefits. They access healthcare support. Social welfare rolls expand. The automatic stabilisers — the very mechanisms designed to cushion economic shocks — kick in, widening the deficit from both sides simultaneously.

Step four — the debt spiral begins. Governments borrow to bridge the gap. But here is the trap: research across 101 economies spanning nearly three decades has established that once public debt crosses approximately 77% of GDP, each additional percentage point of debt actively costs economic growth — roughly 0.017 percentage points annually. The medicine begins to poison the patient. For emerging economies, that threshold falls even lower, to around 64% of GDP.

Step five — interest payments consume the margin. As debt rises, so do interest costs. The United States is already spending nearly $1 trillion annually on debt service — consuming close to one in five tax dollars collected. As displacement erodes the tax base and forces additional borrowing, interest payments crowd out the very discretionary spending — education, infrastructure, retraining — that might otherwise cushion the transition.

The compounding effect means a 10% workforce displacement does not produce a 10% fiscal shortfall. Modelling the full cascade — revenue loss, consumption decline, mandatory spending surge, and debt service acceleration — produces an effective fiscal impact approaching 25% within three to five years. This is the hidden danger: the damage does not arrive as a single shock that triggers a political response. It compounds quietly, year on year, as borrowed money accumulates interest on a shrinking revenue base. A system operating near its thresholds does not absorb that. It breaks.


Developing Nations: The Margin Is Already Gone

If the fiscal arithmetic is alarming for developed nations, for developing economies it is existential — and the timeline is shorter.

The structural vulnerability differs in two critical ways. First, the debt threshold at which trouble begins is lower: research suggests emerging markets face growth-damaging effects above 64% of GDP, against 77% for advanced economies. Several major emerging economies are already uncomfortably close.

Second, and more fundamentally, developing nations have a far thinner formal tax base to begin with. In the Philippines, approximately 76% of employed workers operate in the informal economy, generating little direct tax revenue. The entire fiscal system rests on a narrow band of formal-sector workers — precisely the workers employed in BPO, financial services, and administrative roles that AI is targeting first.

When AI displaces a formal-sector worker in a developing economy, there is no informal fallback for the government. That revenue is simply gone, with no compensating capital gains tax or corporate profit uplift to soften the blow. The OECD's 2026 Economic Survey of the Philippines explicitly identified AI-driven labour market dislocation as a tail risk capable of major macroeconomic disruption. It is one of the few official bodies to say so plainly.


The Political Blindspot

What makes this particularly dangerous is that the political system is architecturally ill-equipped to respond to it.

Democratic governments are designed to react to acute crises — a market crash, a pandemic, a war. They have a poor track record responding to slow-moving structural shifts without a clear triggering event. Japan's experience is the cautionary case: its economy grew at just 1.1% annually between 1991 and 2003 — well below every comparable industrialised nation — while its nominal GDP fell from $5.5 trillion to $4.3 trillion over thirty years.

AI displacement will not arrive as a single shock. It will arrive as a series of quarterly employment reports that look slightly worse than expected, a gradual softening in payroll tax receipts, an uptick in welfare claims that seems manageable in isolation. By the time the fiscal cascade becomes visible in the data, the mandatory spending obligations will have locked in years of commitments that cannot be unwound.

This is not a prediction about a distant future. The displacement is already underway. In the first half of 2025 alone, nearly 78,000 tech job losses were directly attributed to AI. Goldman Sachs estimates 300 million jobs globally face meaningful automation exposure. The IMF puts AI's impact on 40% of all jobs worldwide — rising to 60% in advanced, digitised economies.


What Needs to Happen

The fiscal response required is not modest tinkering. It is structural redesign.

Governments need to begin migrating their revenue base away from labour — toward consumption, capital, and the value generated by AI systems themselves. A Brookings Institution paper published in early 2026 argued for precisely this: modernising consumption tax systems for the digital economy, building administrative capacity to tax AI-generated value, and maintaining flexibility to tax AI-related resource accumulation as systems become the primary drivers of economic value creation.

None of this is politically easy. All of it takes time — typically a decade or more to shift the fundamental architecture of a tax system. Which means the window to act is not in five years. It is now.

The debate about how many jobs AI will destroy is important. But it is secondary to a question governments have barely begun to ask: what happens to the fiscal foundations of the state when it does?

The answer, if you follow the mechanics carefully, is that a 10% problem is never just 10%. And the distance between manageable and broken is far shorter than anyone in power currently believes.


Daniel Horan is Co-Founder & CEO various companies across Europe, Middle East and APAC. Most recently he is founder of RapidHire, an AI-powered hiring platform operating across Southeast Asia. He is writing The Uncomfortable Elite, a book on the fiscal and social consequences of AI-driven labour displacement, forthcoming 2026.


Word count: ~1,200 words

Target outlets: Financial Times Opinion / Bloomberg Opinion / Project Syndicate

Suggested pitch subject line: "The 10% Problem — why AI job displacement breaks national budgets faster than the numbers suggest"

Reference material for this version

Supporting References & Citations

"The 10% Problem" — Daniel Horan

Status: Pre-submission reference document


CLAIM-BY-CLAIM REFERENCE MAP


CLAIM 1

"~60% of federal spending is legally mandated"

Used in: "A Government Is a Fixed-Cost Machine"

Source: US Congressional Budget Office (CBO), The Budget and Economic Outlook: 2024–2034, February 2024.

• Mandatory spending (Social Security, Medicare, Medicaid) = $4.1 trillion = 60.4% of all federal outlays in FY2024

• Adding net interest payments = ~73% of all outlays

• URL: https://www.cbo.gov/publication/59946

Status: ✅ STRONG


CLAIM 2

"Labour accounts for approximately 75% of government tax receipts"

Used in: "Where Government Revenue Actually Comes From"

Source 1: US Congress, Joint Committee on Taxation, Overview of the Federal Tax System in 2024, JCX-6-24.

• Individual income taxes: ~49% of federal revenue

• Payroll taxes: ~36% of federal revenue

• Combined labour share: ~85% (US); OECD average ~75%

Source 2: Korinek, Anton and Lee M. Lockwood, "The Future of Tax Policy: A Public Finance Framework for the Age of AI", Brookings Institution, January 8, 2026.

• Direct quote support: "about three quarters of all US federal tax revenue comes from labour"

• URL: https://www.brookings.edu/articles/future-tax-policy-a-public-finance-framework-for-the-age-of-ai

Status: ✅ STRONG — dual-sourced


CLAIM 3

"Even modest labor displacement could significantly strain public finances"

Used in: "The Cascade — Step 2"

Source: Korinek, Anton and Lee M. Lockwood, "The Future of Tax Policy: A Public Finance Framework for the Age of AI", Brookings Institution, January 8, 2026.

• Exact language from paper: "While the extent and timing remain uncertain, even modest labor displacement could significantly strain public finances at a time when funding for social safety nets may be needed most."

• Authors: Korinek is Professor of Economics, Darden School of Business, University of Virginia; non-resident senior fellow, Brookings Centre on Regulation and Markets. Lockwood is Professor of Economics, University of Virginia.

• Full working paper (53 pages): "Public Finance in the Age of AI: A Primer", same authors, January 2026

• URL: https://www.brookings.edu/articles/future-tax-policy-a-public-finance-framework-for-the-age-of-ai

Status: ✅ STRONG — exact quote confirmed


CLAIM 4

"Public debt above 77% of GDP costs 0.017 percentage points of growth per additional point; emerging market threshold = 64%"

Used in: "The Cascade — Step 4" and "Developing Nations" section

Source: Kumar, Manmohan S. and Jaejoon Woo, "Public Debt and Growth", IMF Working Paper WP/10/174, International Monetary Fund, July 2010.

• Study covered 101 advanced and emerging economies, 1980–2008

• Findings: 77% debt-to-GDP threshold for advanced economies; 64% for emerging markets

• Above threshold: each additional percentage point of debt reduces annual real GDP growth by ~0.017–0.02 percentage points

• URL: https://www.imf.org/external/pubs/ft/wp/2010/wp10174.pdf

Supporting source: Checherita-Westphal, Cristina and Philipp Rother, "The Impact of High Government Debt on Economic Growth and its Channels: An Empirical Investigation for the Euro Area", European Central Bank Working Paper No. 1237, 2012. (Corroborates 90–100% threshold for euro area nations.)

Status: ✅ STRONG — IMF primary source


CLAIM 5

"US spends nearly $1 trillion annually on debt service"

Used in: "The Cascade — Step 4"

Source: Congressional Budget Office, Monthly Budget Review, FY2024.

• Net interest payments FY2024: $892 billion

• Projected FY2025: $1.1 trillion

• As share of revenue: approximately 18–20 cents of every tax dollar collected

Status: ✅ STRONG


CLAIM 6

"Philippines: approximately 76% of workers in the informal economy"

Used in: "Developing Nations" section

Source 1: Philippine Statistics Authority (PSA), 2023 Informal Sector Survey, released 2024.

• Informal sector employment: 76.2% of total employed persons

• Informal sector contribution to GDP: approximately 43%

Source 2: International Labour Organization (ILO), World Employment and Social Outlook, 2023. Corroborates Philippine informal economy figures within 1–2 percentage points.

Note for author: Confirm the PSA figure is from the most recent available survey year. The 76% figure is consistent across 2021–2023 PSA data.

Status: ✅ STRONG — specify survey year in footnote


CLAIM 7

"Goldman Sachs: 300 million jobs globally face meaningful automation exposure"

Used in: "What Needs to Happen"

Source: Briggs, Joseph and Devesh Kodnani, "The Potentially Large Effects of Artificial Intelligence on Economic Growth", Goldman Sachs Global Investment Research, March 26, 2023.

• Key finding: ~300 million full-time equivalent jobs globally exposed to automation

• Two-thirds of US and European jobs exposed to some AI automation; 25–50% of workload automatable

• Widely cited; original report available via Goldman Sachs Research portal

Status: ✅ STRONG


CLAIM 8

"IMF: AI impacts 40% of jobs globally, rising to 60% in advanced economies"

Used in: "What Needs to Happen"

Source: Cesa, Mauro et al., "Gen-AI: Artificial Intelligence and the Future of Work", IMF Staff Discussion Note SDN/2024/001, International Monetary Fund, January 2024.

• 40% of global employment has high AI exposure

• Advanced economies: ~60% of jobs have high AI exposure

• Emerging markets: ~40%; Low-income countries: ~26%

• URL: https://www.imf.org/en/Publications/Staff-Discussion-Notes/Issues/2024/01/14/Gen-AI-Artificial-Intelligence-and-the-Future-of-Work-542379

Status: ✅ STRONG — IMF primary source


CLAIM 9

"Brookings: shift from labour taxation to consumption taxes to maintain fiscal capacity"

Used in: "What Needs to Happen"

Source: Korinek, Anton and Lee M. Lockwood, "The Future of Tax Policy: A Public Finance Framework for the Age of AI", Brookings Institution, January 8, 2026.

Five specific recommendations from the paper:

1. Shift primary revenue source from labour taxes to consumption taxes

2. Modernise consumption tax systems for digital and AI retail services

3. Avoid taxing AI capital assets in the short term (to avoid distorting infrastructure development)

4. Build administrative capacity to tax AI-generated value

5. Maintain flexibility for future adaptation — specifically a tax on AI-related resource accumulation if AI entities become the primary drivers of value creation

Direct quote: "The choice is clear: design tax systems that harness AI's potential for broad-based prosperity, or watch fiscal frameworks buckle under technological change."

• URL: https://www.brookings.edu/articles/future-tax-policy-a-public-finance-framework-for-the-age-of-ai

• Full working paper: "Public Finance in the Age of AI: A Primer", 53 pages, January 2026

Status: ✅ STRONG — exact recommendations confirmed


CLAIM 10

"Japan: GDP grew ~1.1% annually 1991–2003; nominal GDP fell from $5.55T to $4.27T over 30 years"

Used in: "The Political Blindspot"

Source 1: Wikipedia/academic consensus citing multiple sources — "From 1991 to 2003, the Japanese economy grew only 1.14% annually, while the average real growth rate between 2000 and 2010 was about 1%, both well below other industrialised nations."

Source 2: World Bank GDP data (data.worldbank.org) — confirms Japan's annual GDP growth rates 1991–2010, averaging ~1% per annum.

Source 3: Yoshino, Naoyuki, "Japan's Lost Decade: Lessons for Other Economies", ADBI Working Paper 521, Asian Development Bank Institute, 2015. Academic treatment of the stagnation period.

Note for author: The article currently states "10% over two decades / 0.5%/yr" — the more accurate and citable figure is 1.14% annually 1991–2003, or ~1% average 2000–2010. Recommend updating the article to use the 1.14% figure, which is sourced and more precise. The narrative point (political system failed to respond to slow decline) holds either way.

Status: ⚠️ MODERATE — update article figure from "0.5%/yr" to "1.14%/yr" (1991–2003)


CLAIM 11

"OECD 2026 Economic Survey of the Philippines flagged AI-driven labour dislocation as a tail risk"

Used in: "Developing Nations" section

Source: OECD, Economic Survey of the Philippines 2026, OECD Publishing, Paris.

• Referenced in our research session as explicitly identifying AI-driven labour market dislocation as a tail risk capable of major macroeconomic disruption

Note for author: Before submission, verify the exact language used in the OECD survey. The characterisation "tail risk" and "major macroeconomic disruption" may be our paraphrase rather than their direct language. Access via: https://www.oecd.org/en/topics/economy/philippines.html

Status: ⚠️ MODERATE — verify exact OECD language before submission


CLAIM 12

"Fiscal impact approaching 25% within three to five years"

Used in: "The Cascade — Conclusion"

Source: Author's original fiscal cascade model (Daniel Horan, 2026).

Model components:

• Step 1 — Direct labour revenue loss: 10% displacement × 75% labour share = 7.5% of revenue

• Step 2 — Consumption collapse: displaced workers lose 60% of income; Keynesian multiplier 1.5× applied to consumption tax revenue = ~1.5% additional revenue loss

• Step 3 — Mandatory spending surge: displaced workers' wages = 5% of GDP; benefits at 45% of lost wages = 2.25% GDP = ~6.8% of government revenue base

• Step 4 — Debt service escalation: borrowing to cover ~15.8% annual revenue gap at 4% interest; over 3–5 years compounds to ~25% effective fiscal impact

Year 1 impact: ~16%. Year 3–5 impact: ~22–27% depending on interest rates and benefit levels.

Disclosure language for article: "Modelling by the author" or "The author's fiscal cascade model, incorporating direct revenue loss, consumption effects, mandatory spending obligations, and compounding debt service" — methodology available on request.

This is an original analytical contribution and should be presented transparently as such. It is supported directionally by the Brookings Korinek/Lockwood paper and the IMF SDN/2024/001, neither of which model the precise cascade but both confirm the structural vulnerabilities.

Status: 📊 MODEL-DERIVED — disclose as author's original analysis


RECOMMENDED ARTICLE FOOTNOTE FORMAT

For FT or Bloomberg submission, footnotes would read:

1. Congressional Budget Office, Budget and Economic Outlook 2024–2034, February 2024.

2. Joint Committee on Taxation, Overview of the Federal Tax System in 2024, JCX-6-24; Korinek & Lockwood, Brookings Institution, January 2026.

3. Korinek, A. and Lockwood, L.M., "Public Finance in the Age of AI: A Primer", Brookings Working Paper, January 2026.

4. Kumar, M.S. and Woo, J., "Public Debt and Growth", IMF Working Paper WP/10/174, July 2010.

5. Congressional Budget Office, Monthly Budget Review, FY2024.

6. Philippine Statistics Authority, 2023 Informal Sector Survey.

7. IMF Staff Discussion Note SDN/2024/001, "Gen-AI: Artificial Intelligence and the Future of Work", January 2024.

8. Briggs, J. and Kodnani, D., Goldman Sachs Global Investment Research, March 2023.

9. Korinek & Lockwood, op. cit.

10. World Bank GDP Data; Japan's Lost Decade, ADBI Working Paper 521, 2015.

11. OECD, Economic Survey of the Philippines 2026 [verify exact language].

12. Author's analysis. Methodology available on request.


ONE CORRECTION RECOMMENDED FOR THE ARTICLE

Japan growth figure: Change "real GDP grew only 10% over two decades (~0.5% per year)" to:

> "Japan's economy grew at just 1.1% annually between 1991 and 2003 — well below every comparable industrialised nation — while its nominal GDP fell from $5.5 trillion to $4.3 trillion over thirty years."

This is more accurate, more dramatic in the thirty-year framing, and fully citable.


Document prepared: May 2026. Update OECD and PSA citations before final submission.

Join the launch listEmail enquiriesBack to articles