The Federal Reserve Board published a research note finding that anticipated, sustained productivity gains lower 10-year Treasury yields and a common proxy for the equilibrium real policy rate in its model. Contrary to standard macroeconomic theory, the results suggest that optimism about productivity gains from artificial intelligence (AI) cannot explain the recent rise in long-term yields.
The analysis uses vector autoregressive models estimated over 1990–2019 to isolate news about future total factor productivity. Positive productivity news produces a temporary rise in expected short-term interest rates, but a larger, persistent decline in the real term premium more than offsets that increase. Both nominal and real 10-year Treasury yields consequently fall. The expected real short rate five to 10 years ahead, used as a proxy for the equilibrium real federal funds rate, also declines, challenging approaches that tie that equilibrium rate closely to economic growth.
Applying the framework to developments from the fourth quarter of 2022 through the second quarter of 2026, the analysis finds that positive productivity news in 2023–2025 counteracted upward pressure on yields from negative news in 2022. By the second quarter of 2026, the cumulative contribution of productivity news to real 10-year Treasury yields was near zero or slightly negative. Those yields nevertheless rose by 90 basis points between the third quarter of 2022 and the second quarter of 2026, pointing to explanations other than AI productivity optimism.