WASHINGTON — As artificial intelligence reshapes the economy at an unpredictable pace, a prominent former budget official is urging policymakers to start building fiscal safety nets before the disruptions become crises.
Douglas Elmendorf, who led the Congressional Budget Office from 2009 to 2015 and later served as dean of Harvard’s Kennedy School, argues that the uncertainty surrounding AI’s economic impact should not delay preparation. In a new analysis, he warns that the technology could trigger large-scale worker displacement or a dramatic shift in income from labor to capital owners—outcomes that would demand major fiscal policy responses.
“Fiscal institutions cannot be designed overnight,” Elmendorf writes. He points to the slow response to the “China shock” of the 2000s, when trade-related job losses became entrenched before significant federal aid arrived. If AI concentrates wealth and political power among a small elite, he adds, corrective measures could become even harder to enact later.
Elmendorf proposes two key risks that warrant immediate attention. The first is the possibility of widespread job losses. While AI will create new roles, displaced workers may face prolonged unemployment, permanent earnings declines, or exit from the labor force—damage that extends beyond income to health and family stability.
Current federal assistance, such as Trade Adjustment Assistance, is too narrow and burdensome, requiring workers to prove their job loss was trade-related. That model is ill-suited for an AI-driven economy where layoffs stem from a mix of technology, competition, and restructuring. Elmendorf advocates for a modernized adjustment system open to all significantly displaced workers, offering temporary income support, retraining, job-search help, and wage insurance to offset pay cuts when workers take lower-paying jobs.
He acknowledges that design questions remain—eligibility, generosity, and coordination with unemployment insurance—and that evidence on training programs is mixed. But he insists that starting experiments now is essential so that “policymakers should not have to begin designing a response from scratch” if AI triggers major job losses.
The second risk is a sharp rise in the capital share of national income, which would widen wealth gaps and amplify the political influence of asset owners. Traditional remedies—higher taxes on capital income, wealth, inheritances, or consumption—are well understood, and revenue from AI-driven growth could fund worker support. However, Elmendorf suggests a more durable approach: broadening ownership of financial assets.
One option is a sovereign wealth fund that invests on behalf of the public, though it raises tricky questions about financing, governance, and political interference. Another is placing equity stakes in individual accounts, giving citizens a direct claim on capital returns—but design trade-offs between individual control and durable, broadly held ownership remain unresolved.
Elmendorf’s call comes as fiscal pressures mount; the CBO recently raised its deficit forecast for fiscal 2026 by $200 billion, partly due to tariff shortfalls. He stresses that these policy options require further development and testing before they can be deployed at scale. “The prudent policy approach for the risks posed by AI is to begin that work now—in case the responses are needed and before the need becomes urgent,” he concludes.
