The Trump administration has yet to produce a full budget, as required by law, but that is not stopping some top officials from opining about the future course of federal deficits and debt. In particular, they are telling journalists at The Financial Times and elsewhere that the Congressional Budget Office (CBO) and other academic modelers are overly pessimistic about the fiscal outlook because artificial intelligence (AI) is poised to provide a substantial fiscal dividend. It is a risky bet, especially in light of the experience with internet-based technological improvements at the turn of the century.
CBO may have provided some back-up for AI optimism in a May study outlining various deficit and debt scenarios that diverge from the agency’s base case. Among the alternatives modeled was one reflecting a sustained improvement in total factor productivity (TFP).
TFP measures the economic output per unit of combined labor and capital inputs. In effect, it is a measure of the overall efficiency of the non-farm economy. With more rapid TFP growth, the tax base expands with higher profits and wages, which in turn leads to more government revenue.
In its report outlining the alternative projections, CBO did not cite AI as the potential source for higher productivity, but in a previous December 2024 study, it explored what the nascent technology might portend for the economy. The agency reached no firm conclusions but did not dismiss the possibility that AI could prove to be a highly consequential technological breakthrough.
In CBO’s rapid productivity growth scenario, TFP jumps from 1.0 to 1.5 percent per year starting in 2025 and then remains at this elevated level through 2055. If that were to occur, the agency expects federal debt would grow moderately in the coming years and then level off at 113 percent of GDP in 2055.
CBO’s base case forecast offers a more alarming possibility. In that projection, CBO assumes TFP will rise by 1.0 percent per year through 2055, which is broadly consistent with recent experience. With more moderate overall growth rates, budget deficits continue to widen in the coming years as entitlement spending increases outpace growth in federal tax receipts. By 2055, federal debt is estimated to reach 156 percent of GDP, up from just under 100 percent today.
The Trump administration’s incentive to assume an AI-generated fiscal windfall is clear enough. If AI is the spark for a productivity renaissance, higher tax receipts would lessen the need for unpopular spending cuts.
While faster growth would indeed be a welcome development, recent history shows it would be imprudent to assume the broad take-up of AI technology will provide sustained fiscal relief.
Figure 1 tracks TFP growth rates over selected time intervals going back to 1987. It shows that widespread internet adoption had a positive but short-lived effect on productivity.
From 1990 to 2000, the internet was in its infancy, and productivity growth was modest, averaging just 0.9 percent annually. However, as the technology became a more ubiquitous feature in commerce and business operations, productivity improved. From 2000 to 2007, TFP rose at an average annual rate of 1.3 percent.
But then came the financial crash and deep recession of 2007-2009, which upended the global economy and ushered in a period of stagnant growth and minimal wage gains. From 2007 to 2019, TFP rose at an average annual rate of just 0.7 percent. Whatever positive effect the internet was having on productivity during this period was offset by other factors, including an aging population.
In 2023 and 2024, TFP jumped again as the economy emerged from the COVID-induced disruptions of the previous years.
CBO’s base case long-term forecast assumes a reversion back to 1.0 percent annual TFP increases starting in 2025, while the more optimistic TFP scenario assumes growth will not fall back and indeed will stay elevated for a much longer period than was observed during the dot.com cycle twenty-five years ago. Perhaps AI will prove to be a more transformative breakthrough than connecting households and businesses to global information networks, but that is far from certain at this stage.
AI also might generate fiscal costs which wholly or partially offset the benefits of higher productivity. For instance, the technology might alter the distribution of workers across various industries and lead to higher rates of unemployment. If that were to occur, spending on some federal benefit programs, such as unemployment compensation and Medicaid, might increase more rapidly than is projected in CBO’s base case scenario.
With this much uncertainty, elected leaders should be cautious. It would be easier to roll back tax hikes and spending cuts later based on an improved outlook than to enact emergency fiscal restraint in the future when an expected AI dividend never appears.
The post Total Factor Productivity and the Fiscal Outlook appeared first on American Enterprise Institute – AEI.
