When Consumption Fades: VAT base erosion and the fiscal risk of AI-driven unemployment

AI-driven job disruption could weaken consumption and VAT revenues
Uneven AI adoption may concentrate employment losses
Governments must adapt fiscal policy for the AI era

In 2023, value-added tax (VAT) accounted for about one-fifth of total tax revenues collected by OECD countries, roughly 20.5% of government funds (Revenue Statistics 2025, n.d.). This figure is beyond a neutral financial statistic; it represents a delicate pivot point in fiscal policy. According to the OECD, though technological change has contributed to wage inequality in recent decades, recent evidence indicates that advances in artificial intelligence have not yet significantly affected the wage gap or caused widespread loss of wage income. According to the OECD, VAT revenues in OECD countries increased between 2020 and 2022, reaching their highest levels ever as a share of GDP and total taxation in 2022. This suggests that, despite concerns about VAT base erosion linked to employment declines and softened consumer demand, governments have not experienced a direct, mechanical reduction in VAT revenues in recent years.s. Recognizing this risk profoundly changes the setting of policy decisions. VAT can no longer be viewed as a stable source of income but rather as a fluctuating stream closely linked to employment levels and the uneven spread of artificial intelligence (AI) technologies throughout different economies.

VAT Base Erosion as a Fiscal Transmission Channel

Traditionally, discussions around automation focus on labor markets: certain tasks disappear, workers learn new skills, and new jobs are eventually created. However, this perspective ignores critical short- and medium-term fiscal effects. When wages shrink, so does consumer spending. Because consumption taxes like VAT are collected on purchases, any decline in household demand immediately reduces tax revenues. In countries where VAT constitutes a large, reliable portion of overall revenue, vital public services and social safety nets face an early and direct squeeze. The influence varies by region. For instance, European Union countries tend to have higher standard VAT rates and depend more on consumption taxes than some of their peers. Consequently, a given drop in consumer spending hits their public budgets more severely than in countries that rely more heavily on income or corporate taxes (Consumption Tax Trends 2024, n.d.). This reality demands a policy shift: automation cannot be treated solely as a labor adjustment challenge when government budgets might deteriorate well before labor markets stabilize.

Timing increases the complexity of the situation. Businesses are increasingly acting not just on current productivity levels but on expectations of future gains. According to the OECD, employers’ approaches to implementing AI in the workplace often affect workers’ experiences, and surveys suggest that more training and greater employee consultation can help achieve better outcomes during such changes. This link between expected automation and prompt job cuts creates a policy dilemma. If governments respond with tax adjustments only after widespread unemployment becomes apparent, their actions will be belated. Tax revenue can decline within months, but structural fiscal reforms usually take years to implement.

Uneven AI Adoption and the Risk of VAT Base Erosion

According to the latest OECD data, corporate tax revenues have grown even as statutory corporate tax rates have remained stable, suggesting shifts in corporate tax dynamics rather than clear evidence of VAT base erosion. Eurostat data indicate a rapid increase in AI adoption by businesses in 2024, with usage rising from single-digit percentages to double-digit levels in many European countries (Usage of AI technologies increasing in EU enterprises, 2025). This growth is uneven: large companies and specific sectors are adopting AI faster than smaller firms. (AI adoption by small and medium‑sized enterprises, n.d.) This concentration of AI use draws attention to the risk of job losses concentrated in industries that contribute heavily to VAT revenues, such as services, retail, finance, and manufacturing supply chains—all of which fuel substantial consumer spending (Artificial Intelligence-Driven Automation Causes Major Job Losses and Restructuring Globally, 2026). Simultaneously, OECD data confirm that VAT and similar consumption taxes now represent about 20% of total tax revenues in advanced economies (OECD, n.d.).

According to the OECD, the rapid and uneven adoption of AI, combined with advanced economies’ significant reliance on VAT, raises concerns about increased vulnerability to VAT base erosion. To illustrate the budgetary implications, consider a simple and conservative example. Imagine an economy where VAT accounts for 20% of tax revenue and total tax revenue is about 35% of GDP, typical for many OECD countries. (Consumption Tax Trends 2024, 2024) According to the OECD, labour taxes have pushed tax revenues to a record high in 2024, with the average tax-to-GDP ratio across OECD countries reaching 34.1 percent. If automation were to reduce total labor income by 3 percent of GDP over two years due to layoffs and wage pressures, and if household consumption declined proportionally, VAT revenues could also fall, given that a conservative estimate converts around 6 percent of consumer spending into VAT receipts. Receipts would decline by approximately 0.18% of GDP (calculated as 3% reduction times the consumption share times the VAT effective rate) (Consumption Tax Trends 2024, n.d.). According to the OECD, average tax revenues across member countries remained steady in 2023, even as governments faced rising spending pressures, suggesting that a GDP loss equivalent to 0.18 percent may not directly lead to a substantial drop in tax revenue or prompt immediate cuts or borrowing, provided broader tax collection trends remain stable.

In countries with high VAT rates, the fiscal shock could be even greater. For example, if labor income falls by 5% of GDP and consumption decreases slightly more than proportionally—particularly because lower-income households, which tend to spend a higher share of their income, are disproportionately affected—the VAT revenue shortfall could double or even triple relative to the conservative estimate. Companies that reduce their payrolls before realizing productivity improvements worsen this risk by causing fiscal losses to occur earlier than expected. Current research shows not only the anticipatory pattern of layoffs but also the uneven sectoral distribution of AI deployment. This evidence points to an asymmetric fiscal threat: the problem will first emerge where VAT reliance is high, and AI adoption is most intense (Generative AI set to exacerbate regional divide in OECD countries, says first regional analysis on its impact on local job markets, 2024).

Governing the Age of Unemployed Growth

Recognizing VAT base erosion changes how policymakers should respond. First, short-term fiscal buffers grow vital. Temporary cash transfers, flexible VAT rates on essential goods that can be lowered in downturns, and targeted wage subsidies can soften immediate declines in consumption while upholding fairness. These steps buy time and preserve demand, which helps reduce the mechanical drop in VAT revenues. Second, governments should reconsider their tax mix. Growing reliance on property taxes, natural resource rents, or carefully designed digital service taxes can diversify the revenue base rather than overburden consumption taxes. Third, supply-side policies are vital since they mitigate income losses. Accelerated worker retraining tied to actual job transitions, rapid deployment of public employment programs, and incentives for firms to maintain human roles during periods of technological change can lessen the severity of VAT shocks. These practical measures help sustain consumption and the tax base while new technologies spread. Evidence on these approaches comes from successful policy experiments and fiscal modeling in contemporary public finance research (Tax Administration Digitalisation and Digital Transformation Initiatives, n.d.).

Two additional reforms merit serious discussion. The first is a temporary VAT restoration mechanism: automatic stabilizers that boost transfers to low-income households whenever consumption falls below a set, moving threshold, funded through short-term sovereign borrowing. This design supports demand and stabilizes VAT revenues without indefinitely raising tax burdens. The second reform is a medium-term restructuring of the tax base toward wealth and resource-based instruments. For example, windfall taxes or royalty-style levies could be imposed on firms gaining outsized profits from autonomous AI systems. The underlying principle here is to tax in which economic value accumulates when human labor no longer serves as the main indicator of taxable capacity. These reforms contain political or technical challenges, but they shift the fiscal system away from reactive austerity measures toward proactive resilience.

In conclusion, VAT is far from being a mere accounting entry; it is a revenue stream highly sensitive to consumer purchasing and employment levels. The term VAT base erosion captures a pressing policy challenge: fiscal pressure that emerges before labor markets adjust to emerging realities. This risk is particularly acute in countries that depend heavily on VAT and where AI adoption is concentrated—conditions already evident in multiple advanced economies. Authorities must move beyond simplistic long-term tax theories and emphasize short-term fiscal supports, diversification of revenue sources, and active labor market interventions as central components of AI governance. The choice is clear: either accept repeated budget shocks and fragile public services or act decisively now to protect consumption, sustain public investment, and adapt taxation to the evolving foundations of economic value. According to the OECD, some governments have used VAT exemptions, such as Jamaica's 2023 measure exempting imported livestock, to support economic diversification, food security, and employment in targeted sectors.

References

Autor, D., Mindell, D. and Reynolds, E., 2020. The Work of the Future: Building Better Jobs in an Age of Intelligent Machines. Cambridge, MA: MIT Task Force on the Work of the Future.
Brynjolfsson, E., Rock, D. and Syverson, C., 2021. The Productivity J-Curve: How Intangibles Complement General Purpose Technologies. American Economic Journal: Macroeconomics, 13(1), pp.333–372.
Davenport, T.H. and Srinivasan, L., 2026. Companies Are Laying Off Workers Because of AI’s Potential—Not Its Performance. Harvard Business Review.
Eurostat, 2025. Usage of AI Technologies Increasing in EU Enterprises. Luxembourg: European Commission.
Korinek, A. and Lockwood, B., 2023. Future Tax Policy: A Public Finance Framework for the Age of AI. Washington, DC: Brookings Institution.
Korinek, A. and Lockwood, B., 2024. Public Finance in the Age of AI: A Primer. Washington, DC: Brookings Institution.
McKinsey Global Institute, 2025. The State of AI: Global Survey 2025. New York: McKinsey & Company.
Mocuta, S.M., 2026. Weekly Economic Perspectives. State Street Global Advisors.
OECD, 2024a. Consumption Tax Trends 2024. Paris: Organisation for Economic Co-operation and Development.
OECD, 2024b. Generative AI Set to Exacerbate Regional Divide in OECD Countries, Says First Regional Analysis on Its Impact on Local Job Markets. Paris: Organisation for Economic Co-operation and Development.
OECD, 2025a. AI Adoption by Small and Medium-Sized Enterprises. Paris: Organisation for Economic Co-operation and Development.
OECD, 2025b. Revenue Statistics 2025. Paris: Organisation for Economic Co-operation and Development.
OECD, 2025c. Tax Administration Digitalisation and Digital Transformation Initiatives. Paris: Organisation for Economic Co-operation and Development.
OECD AI Policy Observatory, 2026. AI-Driven Automation Causes Significant Job Losses and Restructuring Globally.
Stanford Institute for Human-Centered Artificial Intelligence, 2025. AI Index Report 2025. Stanford University.
State Street Global Advisors, 2026. Weekly Economic Perspectives. Boston: State Street Global Advisors.
Reddit, 2026. How Ironic Will It Be if AI Starts Taking Away Jobs? Discussion thread in r/ArtificialIntelligence.