Fiscal policy will need new tools for uncertain AI future

What’s happening with fiscal policy and AI?

Artificial intelligence could reshape the U.S. economy in profound ways, but scholars, business leaders, and the public aren’t yet sure how. AI might boost productivity while leaving the distribution of income largely intact, or it might trigger sweeping job displacement and a dramatic shift of income toward the wealthiest Americans. Each of these paths would have different implications for the federal budget and would call for different policy responses–but policymakers must set fiscal policy today, without the benefit of knowing what’s ahead.  

A new working paper by Karen Dynan, professor of the practice of economics and public policy at HKS, and Douglas Elmendorf, Lucius N. Littauer Professor of Public Policy at HKS, along with Louise Sheiner of the Brookings Institution, examines four illustrative scenarios for how AI might transform the economy and traces what each would mean for the federal debt and economic policy.  

What does the research say?

The authors construct four scenarios that combine assumptions about productivity growth, job displacement, and how income gains are distributed, both between capital and labor and across the income spectrum. They stress that these scenarios are illustrative rather than predictive, given the volatile nature of the AI landscape.

The authors identify several key possibilities.  

In the first, most optimistic scenario, AI drives productivity, income growth follows, and benefits everyone across the economy in proportion to existing income levels across the economy.

The second scenario keeps the same productivity boost, income growth follows, but nearly all income flows to the top income earners. Growth accelerates just as before, but the disparity gap between high-and low-earners widens.  

Layered onto that inequality, the third scenario introduces disruption into the flow of the job market. Though AI will create opportunities, it will also eliminate the need for some jobs. This may result in an increase in the unemployment rate as workers need time to find new roles. Some workers may step away from the workforce entirely.

In the most disruptive fourth scenario, both productivity and job loss peak. Long-term unemployment rises, more people leave their jobs, and all of the additional income goes to those who own capital. Job loss from AI carries ramifications beyond lost income, including documented links between job displacement and higher mortality, worse mental health, and an eroded sense of purpose.

How can policymakers respond?  

According to Dynan, Elmendorf, and Sheiner, policymakers have a wide menu of possible responses depending on which scenario unfolds. Options include expanding financial support and job retraining for displaced workers, making taxes and government benefits more generous for lower earners, or more far-reaching changes like taxing wealth and investments at higher rates or giving the public a stake in business ownership, for instance, through a government investment fund or individual accounts.

The authors note that a “wait and see” approach—delaying major policy changes until AI’s full economic effects become apparent—has the advantage of allowing for more precise responses. But they also point out a real risk to waiting, as understanding often develops slowly, and policy adjustments take even longer. By the time disruptive effects are clear, entrenched interests may hinder some policy responses.  

As an alternative, the authors describe an “insurance” approach, adopting modest, adaptable policies now, such as a scaled-down version of trade adjustment assistance or small public equity purchases, that could be expanded quickly if more disruptive scenarios materialize, but would impose limited costs if they do not. Whether the benefits of this approach exceed the costs, the authors note, is difficult to evaluate, given how uncertain AI’s economic path remains. 

—
Photo by Brendan Smialowski / AFP via Getty Images