Basil Halperin on Macroeconomic Policy in an Age of Transformative AI

Episode Summary

Executive Summary: David Beckworth and Basil Halperin discuss viral AI dystopia narratives, arguing that transformative AI would likely raise real interest rates via higher growth and lower savings, but current market rates do not yet imply such a regime. They then turn to Halperin’s menu-cost research, which suggests monetary policy should “look through” supply shocks and target nominal income more than inflation.

Main Topics: Viral AI dystopia and the “Macro Memo from June 2028” (Priority: 5/5): They revisit a widely read essay predicting AI-driven layoffs, collapsing labor income, weak demand, and a “ghost GDP” world, using it as a springboard to discuss how compelling narratives can shape markets and public debate. Interest rates as a signal of transformative AI (Priority: 5/5): Halperin explains that if markets expected a near-term AI-led growth explosion, real interest rates would rise sharply because faster future income and lower near-term saving both increase rates. Why the market signal suggests AI has not yet reached transformative scale (Priority: 4/5): He argues that current long-term real rates are elevated relative to the COVID era but still far from the levels implied by something like 30% GDP growth, so financial markets are not pricing a full transformative-AI scenario. Policy response to AI-driven transition (Priority: 4/5): The conversation explores redistribution, universal basic income, and fiscal adjustment as the main tools for handling AI-created dislocations, with monetary policy focused mainly on avoiding mistakes rather than solving distributional conflict. Menu costs, sticky prices, and optimal monetary policy (Priority: 5/5): Halperin summarizes his research showing that with menu costs, it is inefficient for all firms to adjust prices after one firm-specific shock; instead, monetary policy should allow inflation to move countercyclically, closer to nominal GDP targeting than strict inflation targeting. Broader nominal rigidities and the case for NGDP targeting (Priority: 4/5): They discuss sticky wages, sticky information, debt contracts, and bounded rationality, noting that many distinct frictions point toward nominal GDP or nominal income targeting as an eclectically optimal policy framework. Long-run technological change, energy shocks, and flexible prices (Priority: 3/5): They distinguish between sticky and flexible-price sectors, noting that energy shocks may warrant “looking through” because they are flexible-price shocks, while AI and technology may eventually reduce nominal rigidities or even reshape transactions entirely.

Key Arguments: Narrative-driven predictions about AI can be vivid but fail general-equilibrium consistency; math is needed to ensure that income, output, and demand identities still hold together. If transformative AI really implied a near-term jump toward something like 30% GDP growth, long-term real interest rates should already be much higher; the absence of such a signal suggests markets are not there yet. Both rapid future growth and existential-risk scenarios push real rates up because they reduce the incentive to save today (consumption smoothing). Higher AI-related capital demand from firms building data centers and compute clusters can also raise rates through the demand side. Distributional consequences of AI are likely to require fiscal redistribution, not just monetary policy; central banks should mainly avoid destabilizing the transition. Under menu costs, forcing all firms to adjust nominal prices after a sector-specific shock is wasteful; letting the directly affected firms adjust while monetary policy stabilizes nominal spending is more efficient. Many nominal frictions—sticky wages, sticky information, nominal debt, and possibly bounded rationality—point to nominal GDP or nominal income targeting as a practical policy rule. Energy shocks are different from broad demand shocks because energy prices are relatively flexible, so monetary policy should mostly look through them rather than fully offset them. Transformative AI could eventually make some rigidities less important via electronic shelf labels, AI agents, or even frictionless/asset-based payments, but human behavioral frictions may persist much longer. Debt sustainability under transformative AI is ambiguous: faster growth raises the tax base, but higher interest rates raise rollover costs, and the net effect depends on the elasticity of intertemporal substitution.

Data Points: Transformative AI benchmark: ~10x speedup in GDP growth / about 30% GDP growth - Halperin’s benchmark for truly transformative AI in the conversation and related paper Long-term real interest rate: 2.75% - The 30-year real rate discussed as elevated but still within historical norms Stock market duration: 10–20 years - Used to illustrate that financial markets are inherently forward-looking Anthropic revenue growth: 10x a year - Cited as an example of rapid AI-company growth even absent economy-wide transformation U.S. debt-to-GDP ratio: 90% - Used in discussion of how transformative AI might affect fiscal sustainability Average Treasury maturity: about 6 years - Illustrates how quickly higher interest rates would affect federal rollover costs Elasticity of intertemporal substitution estimate: ~0.7 - Mentioned as a micro-based estimate suggesting interest rates could rise more than growth in some scenarios Late-1990s GDP growth: about 4% - A comparison point showing how unusual 5% growth would be

Pivotal Quotes: "Doing macroeconomics in words is hard." — Basil Halperin: Explaining why the viral AI essay’s logic needs formal modeling to remain internally consistent "What monetary policy can do is ensure that it doesn't screw up the economy in response to shocks." — Basil Halperin: Summarizing his view that central banks should avoid destabilizing nominal rigidities rather than solve redistribution "Eclectically optimal" — Basil Halperin: His characterization of nominal GDP targeting as a practical policy rule supported by many different frictions

Implications: Markets do not yet price a near-term AI super-boom, but if it arrives, rates, debt dynamics, and distributional politics will shift sharply. For central banks, the practical lesson is to avoid overreacting to supply shocks and to keep nominal spending on a stable path.