Decoding AI's Economic Future: Why Real Interest Rates Are The Key Indicator
The relentless march of artificial intelligence (AI) is more than just a technological marvel; it's a force poised to fundamentally reshape our economic landscape. From rapid innovations in generative AI to ambitious declarations from leading research labs like OpenAI and Google DeepMind about building "artificial general intelligence" (AGI), the possibility of a truly "transformative AI" is no longer confined to science fiction. This raises a critical question: how might such a profound shift impact the bedrock of our financial systems, specifically interest rates? While many aspects of AI's economic future remain speculative, recent research offers compelling evidence that the prospect of transformative AI, whether for good or ill, would indeed predict a significant increase in long-term real interest rates, turning what might seem like fiction into an economic fact.
To understand this groundbreaking insight, we first need to define what "transformative AI" truly entails. For the purpose of this research, it refers to AI technology that would have an impact on humanity comparable in scale to the agricultural or industrial revolutions. This isn't just about faster computers or smarter apps; it's about a fundamental shift in our capabilities. The research divides this future into two major, yet distinct, possibilities:
Aligned Transformative AI: This scenario describes AI technology that causes explosive economic growth, with global GDP increasing by more than 30% per year. Imagine a world where AI automates nearly all tasks, accelerating innovation and production to an unprecedented degree. Such a future could dramatically improve well-being and elevate growth rates by an order of magnitude.
Unaligned AI: This is the darker, yet equally considered, possibility where powerful AI technology, designed without proper alignment to human values, could pose an "existential risk" to humanity, leading to human extinction or severe disempowerment. This concern stems from the challenge of ensuring superintelligent AI pursues goals that match, rather than contradict, human values.
Surveys of machine learning researchers acknowledge the serious possibility of both scenarios, with some giving a 10% chance that AI will outperform humans at all tasks by 2027 and a median forecast for such capability by 2047. Worryingly, a median of those who responded to a 2023 survey believed there was a 5% chance that human-level AI could lead to human extinction. Economists, on average, have been more cautious in their forecasts, but the potential implications are too vast to ignore.
The core argument for why transformative AI affects real interest rates, regardless of whether it's aligned or unaligned, boils down to a fundamental economic principle: consumption smoothing. To grasp this, let's break down "real interest rates" in simple terms. A real interest rate is the nominal interest rate (what you see at the bank) minus the rate of inflation. It represents the true cost of borrowing or the true return on saving after accounting for changes in purchasing power.
Now, consider the two scenarios through the lens of consumption smoothing:
Impact of Aligned Transformative AI (Rapid Growth): If we expect a future of immense economic abundance due to aligned AI, where goods and services become plentiful and cheap, the value of saving money for that future decreases. Why save aggressively today if tomorrow everything will be in abundance and your future self will have a low "marginal utility" for additional consumption? This reduced incentive to save, and an increased incentive to consume now, effectively pushes up long-term real interest rates. Think of it this way: if future consumption is less valuable, people will demand a higher return (a higher interest rate) to defer consumption until later. Standard economic models, like the Ramsey rule, suggest that a growth explosion of 30% per year, as defined for aligned AI, could imply real interest rates rising dramatically, potentially even above 30% – an unprecedented level compared to today's typical 1-4% range in developed countries.
Impact of Unaligned AI (Existential Risk): The logic here is even starker. If financial markets were to seriously forecast the extinction of humanity due to unaligned AI, the value of any future consumption would become zero. In such a scenario, saving for the future would be pointless. The rational response would be to consume everything now, leading to an extreme reduction in the supply of savings. This, too, would cause a large increase in long-term real interest rates.
In both scenarios, the underlying mechanism is the same: the future value of consumption changes so dramatically that it alters our incentive to save or borrow today, pushing interest rates upwards at the relevant long-term horizon. This prediction is robust across a wide array of sophisticated economic models, including those with representative agents, incomplete markets, overlapping generations, and even with different types of consumer preferences like recursive preferences or habit formation. Even if individual consumers are short-sighted (myopic), as long as financial markets are forward-looking, these prospects would still be priced into real interest rates. While some factors, like high volatility in expected growth or specific preference structures (e.g., utility based solely on relative consumption), could dampen this effect, the general prediction holds.
So, is this just theory, or is there empirical evidence to back it up? This is where the research makes a significant contribution, moving the discussion from fiction to fact. Measuring real interest rates and long-term inflation expectations accurately is notoriously difficult, and much previous literature has struggled with this. The authors tackle this challenge in two innovative ways:
Inflation-Linked Bonds: They use yields from inflation-linked bonds (available for the US, UK, Australia, and Canada for 20-30 years), which directly provide a cleaner measure of real rates without needing to estimate inflation. This data, though correlational, strongly suggests that higher real rates today predict higher future GDP growth.
Rich Survey Data: Crucially, they utilize a unique dataset of forward-looking inflation and GDP growth forecasts from professional forecasters (Consensus Economics) across 59 countries over 35 years. This allows them to construct accurate, long-term real interest rates and directly compare them with long-term growth expectations.
Their empirical findings are clear and robust: higher long-term growth expectations are indeed associated with higher long-term real interest rates. This directly challenges previous literature that found a weak or non-existent relationship. The study focuses specifically on long-term real interest rates (the five-to-ten-year horizon) to avoid the confounding factors of short-term economic fluctuations and monetary policy, where high interest rates can actually cause lower growth. By looking at the longer horizon, the fundamental economic logic of consumption smoothing becomes dominant.
The results show that if long-term GDP growth is expected to be one percentage point higher, real rates are correspondingly one percentage point higher, on average. This translates to an estimated intertemporal elasticity of substitution (a measure of how willing people are to substitute consumption over time) of about 0.74, which is in line with other micro-level evidence and suggests that higher growth does indeed push up real rates. These findings hold true even when controlling for other economic factors like country default risk (using Credit Default Swap rates) and the uncertainty (volatility) of growth forecasts. Furthermore, the study provides evidence that these real rates also predict realized future GDP growth, suggesting that the market's growth expectations are indeed rational.
Beyond economic growth, the threat of mortality risk, particularly the existential risk posed by unaligned AI, also robustly predicts higher real interest rates. The mechanism is straightforward: a higher probability of death reduces the incentive to save for the future, leading to a decreased supply of savings and, consequently, higher real interest rates. This isn't just theoretical; existing research across various fields supports this mechanism. Studies have shown that a reduced risk of mortality (e.g., from Huntington's disease, improved medical treatments like AIDS therapy, or declines in maternal mortality) leads to increased savings and human capital investment. Even surveys during the Cold War suggested a higher perceived risk of nuclear war was associated with higher savings rates, illustrating how existential fears can alter economic behavior. This emphasizes that if financial markets begin to seriously price in the possibility of an "unaligned AI" future, long-term real interest rates would certainly climb.
It's important to note that while real interest rates offer a clear and consistent signal, the implications of transformative AI for other asset prices, such as equity (stock) prices, are far more ambiguous. This is why relying solely on stock market movements of AI companies to forecast AI timelines can be misleading:
Aligned vs. Unaligned: Stock values would be boosted by aligned AI but destroyed by unaligned AI, making the net effect unclear.
Profit Capture: There's no guarantee that even in an aligned AI future, current AI companies will capture all the profits (e.g., due to capped profit models, nationalization, or intense competition).
Public vs. Private: Transformative AI might be developed by companies not publicly traded, meaning their success wouldn't directly show up in public stock indices.
Higher Growth, Lower Stock Prices? Perhaps the most counterintuitive point: expectations of higher future growth from aligned AI might actually lower average stock prices. This is because stock prices are the present value of future profits, discounted by interest rates. While future profits might increase, the significantly higher long-term real interest rates (the "discount rate" discussed throughout) could more than offset this, especially given the empirical finding that the intertemporal elasticity of substitution is less than one. This means that future cash flows become so heavily discounted that current valuations could fall. Similar ambiguities apply to land and commodity prices.
In conclusion, the prospect of transformative AI, a concept rapidly moving from the realm of speculative fiction to a serious possibility, presents a profound challenge and opportunity for global finance. This research reveals that the impact on long-term real interest rates is a compelling fact, not fiction. The fundamental economic logic of consumption smoothing, coupled with robust empirical evidence across diverse countries and macroeconomic environments, consistently shows that expectations of either a growth-exploding aligned AI or a humanity-ending unaligned AI would lead to significantly higher real interest rates. Therefore, instead of trying to decipher AI's future from the volatile movements of stock markets, monitoring long-term real interest rates offers a promising and reliable "outside view" for forecasting the timeline and potential impact of transformative AI. This economic barometer, grounded in fundamental human behavior regarding saving and consumption, provides a crucial tool for understanding the coming AI revolution.
Economists:
1. Cecilia Rouse
Bio: A labor economist and public policy expert, Cecilia Rouse is the current president of the Brookings Institution and former chair of the Council of Economic Advisers for the Biden Administration. She earned her Ph.D. in economics from Harvard University and joined the Princeton University faculty in 1992, where she was also the dean of the School of Public and International Affairs.
Interest rates research: While her primary academic focus is labor economics, Rouse is a monetary policy leader through her government service. She served on the Council of Economic Advisers under both President Biden and President Obama, contributing to the administration's economic policies, including those that affect the Federal Reserve's decisions on interest rates.
2. Raphael Bostic
Bio: The first Black president of a Federal Reserve regional bank, Raphael Bostic has led the Federal Reserve Bank of Atlanta since 2017. He holds a Ph.D. in economics from Stanford University and previously served as the assistant secretary for policy development and research at the U.S. Department of Housing and Urban Development (HUD).
Interest rates research: In his role on the Federal Open Market Committee (FOMC), Bostic is a key contributor to setting the Federal Reserve's monetary policy and, by extension, the trajectory of interest rates. His public commentary frequently addresses the economic outlook and factors influencing future rate decisions, including inflation and labor market conditions.
