Creative Destruction Gets Its Nobel

Why do some economies sustain growth while others stagnate? Joel Mokyr, Philippe Aghion, and Peter Howitt spent decades answering that question, and on October 14, 2025, they received the Nobel Prize in Economics for their work on technology-driven growth. Their answer was innovation, *creative destruction*, and the institutional frameworks that enable both. The old must die for the new to be born. The laureates mapped how this happens, and what prevents it. Recognition came as the world grappled with three interrelated puzzles. How AI might transform labor markets, how climate change would reshape economic projections, and whether the era of globalization was ending. Their frameworks, developed to explain the past, now face tests they were not designed for. ## The 2025 Nobel Prize Aghion and Howitt's 1992 paper "A Model of Growth Through Creative Destruction" established the workhorse model of *endogenous growth theory*.[^1] It formalizes Schumpeter's insight that innovation destroys existing economic structures while creating new ones. Firms invest in R&D to acquire patents, hoping to earn monopoly profits before the next innovation renders their products obsolete. In their framework, the economy's steady-state growth rate depends on how often innovations arrive, how much effort goes into producing them, and how large each improvement is: $$g = \lambda \cdot \phi \cdot \ln(\gamma)$$ Here $\lambda$ is the arrival rate of innovations, $\phi$ is the fraction of labor devoted to R&D, and $\gamma$ measures the step size of quality improvements. The logarithm captures diminishing returns: doubling quality matters more when starting from a low base. The key insight is that growth arises endogenously from profit-seeking innovation, not from some external force. Mokyr identified the conditions necessary for sustained growth: a feedback channel between theoretical and practical knowledge, mechanical competence for developing and producing innovations, and institutions open enough to reduce resistance to change.[^2] His historical work traced how these conditions emerged in Europe and not elsewhere, shaping the divergence in living standards we observe today. ## Working Hours Across the World The laureates mapped creative destruction across centuries. Their framework explains why some societies escaped poverty while others remained trapped, why innovation clusters in certain places and times. But explaining history is easier than predicting the future. The datasets released in 2025 offered something new: the chance to test whether these frameworks hold under conditions the laureates never anticipated. One variable the growth models left underspecified was how societies balance productivity against leisure. Amory Gethin and Emmanuel Saez constructed the most comprehensive database on working hours to date, covering 160 countries and 97 percent of the world's population.[^3] Their data combine labor force surveys from multiple sources to enable cross-country comparisons and long-run analysis. Their database reveals substantial variation in working hours across countries, even at similar income levels. Cultural factors, labor regulations, and economic structure all contribute. The data also document a secular decline in working hours in developed economies over the twentieth century, a trend that has slowed or reversed in some countries recently. Contrast is stark. Germans and French work roughly 1,400 hours annually. Americans and South Koreans exceed 1,800. This new dataset allows researchers to test whether these gaps reflect preferences, regulations, or measurement artifacts. Understanding working hours is essential for interpreting productivity statistics, assessing living standards, and designing labor market policy. GDP per capita can mislead about welfare when hours differ substantially. ## Corporate Investment Puzzles Working hours shape what societies produce. Corporate investment shapes what they build next. The labor data revealed stark divergence across countries at similar income levels, differences that productivity models struggle to explain. A parallel puzzle emerged in capital allocation, where firms seemed to reject projects that should create value. Understanding why required looking beyond theory into the messy reality of how decisions actually get made. Standard finance theory prescribes discounting future cash flows at rates reflecting risk and the time value of money. A project's *net present value* (NPV) sums its cash flows, each discounted by how far in the future it occurs: $$NPV = \sum_{t=0}^{T} \frac{CF_t}{(1+r)^t}$$ Here $CF_t$ is the cash flow at time $t$ and $r$ is the discount rate. The formula works because a dollar today is worth more than a dollar tomorrow, both because of impatience and because today's dollar can be invested. Projects with positive NPV create value and should be accepted. The puzzle is not the formula but the inputs. New research analyzing hand-collected data from SEC filings, investor presentations, and CFO surveys found that firms consistently use *discount rates* 3-5 percentage points above what theory prescribes.[^4] These inflated hurdle rates lead companies to reject projects that should create value. The discrepancy between theory and practice has implications for understanding aggregate investment behavior. If firms systematically underinvest due to behavioral biases, macroeconomic models that assume rational capital allocation may mislead. ## AI and the Labor Market If firms systematically underinvest due to inflated hurdle rates, macroeconomic models miss something fundamental about human decision-making. Theory predicts one thing; behavior delivers another. That gap between prescription and practice mattered especially in 2025, when economists faced a challenge their models had never encountered: estimating what happens when machines can reason. Daron Acemoglu's "The Simple Macroeconomics of AI," published in *Economic Policy*, provided the most skeptical assessment from a leading economist. His *task-based framework* distinguishes between automation, replacing labor, and augmentation, enhancing labor productivity. His headline finding is that AI advances are likely to increase *total factor productivity* (TFP) by no more than 0.66% over ten years. Meaningful, but far from transformative. Other economists project larger impacts, ranging to 2-3% TFP gains. The gap reflects differing assumptions about task automation shares, diffusion speed, and complementary investments in reorganization and training. Acemoglu's modest estimate arises from a structural constraint. AI, like previous automation technologies, affects only a subset of tasks. Most economic activity involves tasks that AI cannot currently perform: physical manipulation, nuanced human interaction, creative judgment under uncertainty. *Hulten's theorem*[^5] limits macroeconomic impact to the productivity gain in automated tasks multiplied by those tasks' share of total output. The intuition: economy-wide gains equal sector productivity gains weighted by sector size. AI transforming a small slice of tasks produces small aggregate effects. (See our CS & AI coverage for technical details on current capabilities.) Other research documented more immediate effects. Analysis of job postings found a 24% decrease in generative AI-exposed skills among highly automatable jobs, alongside a 15% increase in such skills for augmentable positions. The labor market is bifurcating. AI complements some workers while substituting for others. ## Inflation's Aftermath Acemoglu's skepticism anchored one side of the AI debate; optimists projected transformative gains. The disagreement reflected how little economists know about productivity shifts as they happen. Models can only be tested against outcomes, and outcomes take time. That uncertainty had a recent precedent: the inflation surge that no model predicted, and whose aftermath continued to puzzle central banks well into 2025. That surge receded, but economists continued debating its causes and consequences. Central bank forecasting came under scrutiny. A formal evaluation of Bank of England inflation forecasts found that despite large forecast errors during the surge, predictions remained accurate relative to common benchmarks.[^6] The errors reflected the exceptional nature of the pandemic and energy shocks rather than systematic bias. Questions of *inflation persistence* dominated monetary policy discussions. Would inflation settle back to 2% targets, or had the high-inflation episode changed expectations and wage-setting behavior? Evidence through 2025 suggested that anchored expectations held, with long-term inflation expectations remaining near targets despite the temporary surge. This revived debates about central bank frameworks. Some argued for higher inflation targets. Others advocated *nominal GDP targeting*. Still others called for returning to the pre-2008 consensus with humility about model limitations. ## Climate Economics Updated The inflation debates revealed how much damage misprediction causes. Models failed, forecasts missed, and central banks scrambled to catch up with a reality they had not anticipated. A longer-running failure of economic coordination cast a larger shadow: climate policy, where the gap between optimal and actual remains an order of magnitude, and where the costs of error compound not over quarters but across generations. William Nordhaus, the 2018 Nobel laureate, released updated estimates from the *DICE-2023* model, his Dynamic Integrated Climate-Economy framework.[^7] The revised *damage function* projects that 3°C warming would reduce global GDP by approximately 2.1%, rising to 8.5% at 6°C. These estimates remain controversial. Critics argue that *integrated assessment models* understate tail risks including conflict, mass migration, ecosystem collapse, and non-market harms not captured in GDP. The baseline estimate of the *social cost of carbon*, the economic damage from emitting one additional tonne of CO₂, is $66 per tonne. This aligns with other recent estimates when differences in discounting are accounted for. The policy implications are stark. Current policies will cause global mean temperature to pass the 1.5°C threshold later this decade and exceed the Paris Accord's 2°C goal by mid-century. Achieving Paris goals would require a global *carbon price* of approximately $80 per tonne. The World Bank estimates the current effective global price at roughly $3 per tonne (as of Q3 2025). The stakes are substantial. The net present value of benefits from optimal climate policy is estimated at $120 trillion. The gap between optimal and actual policy represents one of the largest failures of collective action in history. ## Wealth Inequality Dynamics Nordhaus's updated estimates put numbers on what climate inaction costs. The $120 trillion in foregone benefits from optimal policy represents a failure of collective action unprecedented in scope, a slow-motion catastrophe that economics can measure but politics cannot prevent. Climate is not the only domain where costs compound quietly. Wealth, too, accumulates and concentrates through mechanisms that operate over generations, visible only when someone bothers to measure. *World Inequality Lab* released the first comprehensive global database on wealth accumulation, covering all world regions from 1800 to 2025.[^8] The findings update and extend Thomas Piketty's work on capital accumulation. Global *wealth-income ratios* have risen dramatically from approximately 390% of world net domestic product in 1980 to over 625% in 2025. Wealth has risen much faster than income. New empirical work tested Piketty's *r > g* hypothesis: that a persistent gap between returns on capital (r) and economic growth (g) drives wealth concentration. Using data from 16 developed countries spanning 1870-2020, researchers found support for the mechanism but also identified moderating factors. More egalitarian political institutions and left-leaning governments appear to mitigate the effect, while the repeal of wealth taxes amplifies it. In 2024, the G20 summit formally discussed a global minimum tax on billionaires, a proposal that would have seemed radical a decade ago but is now part of mainstream policy debate. Though the proposal was not adopted, its serious consideration marks a shift in acceptable policy options. ## DeFi's Hidden Vulnerabilities The r > g debate placed wealth concentration at the center of policy discussions. Billionaire taxation moved from radical proposal to G20 agenda item. But while economists debated redistribution within the existing financial system, a parallel system was growing outside their models entirely. Decentralized finance promised to distribute access rather than concentrate it. Whether it delivered on that promise, or merely distributed new forms of risk, became a question researchers could finally answer in 2025. A comprehensive framework distinguished three categories of DeFi protocols: *liquidity pools*, pegged and synthetic tokens, and *aggregator protocols*. Each carries distinct risks. Liquidity pools face *impermanent loss* from price divergence. Pegged tokens (including *stablecoins*) risk depegging when collateral becomes insufficient. Aggregators inherit and amplify the risks of underlying protocols. A concept of *crosstagion*[^9] emerged to describe bidirectional contagion between traditional finance (TradFi) and DeFi. A liquidity crisis in either system can cascade into the other through shared participants and correlated positions. The 2022 collapse of FTX and its contagion through crypto markets demonstrated this dynamic in practice. DeFi's capital efficiency remains constrained by the absence of effective credit risk assessment. On-chain credit scoring attempts to address this, enabling dynamic adjustment of loan-to-value ratios based on wallet behavior. But reputation systems in pseudonymous networks remain primitive compared to traditional credit markets. ## Central Bank Digital Currencies Advance DeFi revealed both the appetite for financial innovation and the risks that come with it. Crosstagion between traditional and decentralized systems showed that these worlds were not as separate as proponents claimed. Central banks took note. Their response was not to regulate away the innovation but to build their own version: digital currencies that might capture the benefits of programmable money while retaining institutional control. *Central bank digital currencies* (CBDCs)[^10] moved from research to implementation across much of the world in 2025. As of November 2025, 137 countries representing 98% of global GDP are exploring CBDCs, up from just 35 in May 2020. Forty-nine pilot projects are underway. Three countries have fully launched digital currencies, the Bahamas, Jamaica, and Nigeria. China's *e-CNY* pilot remains the most advanced among major economies, now covering 26 regions in 17 provinces. The pilot extended to Hong Kong, creating the world's first link between a *faster payments system* and a CBDC. India's *e-rupee* is the second-largest pilot, with digital currency in circulation rising 334% to $122 million by March 2025. Europe's Central Bank is moving toward a *digital euro*, with a potential development phase beginning in November 2025. Sixteen G20 countries are in active development or pilot phases. America remains an outlier. The House passed legislation that would prohibit the Federal Reserve from testing or implementing a CBDC, though the bill awaits Senate action. The divergence in approaches between major economies may fragment the international monetary system. ## Geoeconomic Fragmentation America's resistance to CBDCs while China expanded the e-CNY illustrated something larger than monetary policy divergence. The dollar-based system that had organized global finance for decades was fragmenting along geopolitical lines. Trade restrictions tripled since 2019. Supply chains realigned toward allies. The costs of this unwinding, concentrated among the poorest countries, were just beginning to be measured. IMF documented accelerating *geoeconomic fragmentation*.[^11] *Strategic decoupling*, the deliberate unwinding of economic ties between geopolitical blocs, would impose permanent GDP losses of 0.3% globally, equivalent to Norway's annual output. But the burden falls unevenly. Low-income countries would lose more than 4% of GDP, deepening debt risks and food insecurity. Fragmentation is more costly now than during the Cold War because the starting point is more integrated. Goods trade as a share of GDP was 16% at the Cold War's onset. It is 45% today. Unwinding these linkages destroys more value. Trade flows are already realigning. While global trade has declined since 2022, it has fallen less between geopolitically aligned countries than between divergent ones. Companies are *friend-shoring*, relocating supply chains to allies at the cost of efficiency. The implications for the dollar-based international monetary system remain unclear. IMF notes that even modest fragmentation reduces the appeal of reserve currencies by limiting their usefulness in international transactions. Whether alternative systems will emerge remains speculative. ## Frameworks Under Strain The laureates received their prize for explaining how creative destruction drives growth. Their framework was built from centuries of historical data: steam engines and railroads, electricity and automobiles, computers and the internet. Each wave destroyed old industries while creating new ones. The pattern held for two hundred years. The question is whether it holds for the next twenty. AI may be different. Acemoglu's skepticism rests on current capabilities, and current capabilities are changing faster than any previous technology. The models may not be wrong. They may simply be too slow to capture a phenomenon that rewrites its own assumptions. Climate economics captures a different failure. At $3 per tonne versus a needed $80, the carbon pricing gap represents a market distortion of historic proportions. The models know what should happen. Politics ensures it does not. The fragmentation of trade and the divergence in monetary systems may be reshaping the international order in ways that will only become clear over decades. Economists are documenting changes whose significance they cannot yet assess. ## Models and Their Limits Productivity growth has slowed in developed economies despite continued technological innovation. The creative destruction framework predicts faster growth as innovation accelerates, yet the numbers move the other way. Either the models miss something fundamental, or the measurements fail to capture what matters. AI compounds this uncertainty. Acemoglu's modest projections assume current capabilities, but recursive improvement could change the calculus entirely. The economics of exponentials is unforgiving: wrong by a factor of two today, wrong by orders of magnitude in a decade. Climate policy lags climate science by decades and may never catch up. The DICE model projects substantial losses from delay, but political economy continues to favor inaction. Geoeconomic fragmentation may deepen or reverse; history offers precedents for both, and which applies depends on choices not yet made. The field enters 2026 with frameworks for understanding growth, inequality, and disruption. Those frameworks were built from two centuries of data in a world where change was fast but directional, where disruption destroyed old industries while creating new ones, where global integration seemed irreversible. The theories were built for a different world. That world may be ending. --- **Citations**: [1] Aghion, P., and Howitt, P. "A Model of Growth Through Creative Destruction." Econometrica, 1992. [2] Mokyr, J. "A Culture of Growth: The Origins of the Modern Economy." Princeton University Press, 2016. [3] "Working Papers." National Bureau of Economic Research, 2025. [4] "Working Papers." National Bureau of Economic Research, 2025. [5] Acemoglu, D. "The Simple Macroeconomics of AI." Economic Policy, Volume 40(121), pages 13-58, 2025. [6] Coroneo, L. "Forecasting for monetary policy." arXiv:2501.07386, January 2025. [7] Barrage, L., and Nordhaus, W. "Policies, Projections, and the Social Cost of Carbon: Results from the DICE-2023 Model." PNAS, 2024. [8] "Global Wealth Accumulation and Ownership Patterns, 1800-2025." World Inequality Lab Working Paper 2025/22. [9] Aufiero, S., et al. "Mapping Microscopic and Systemic Risks in TradFi and DeFi." arXiv:2508.12007, August 2025. [10] "Central Bank Digital Currency: Progress and Further Considerations." IMF Policy Paper, 2024-2025. [11] "Geoeconomic Fragmentation and the Future of Multilateralism." IMF Staff Discussion Notes, 2023. **Footnotes**: [^1]: Endogenous growth theory models growth as arising from within the economy, through deliberate investment in R&D and human capital, rather than from exogenous technological progress assumed in earlier models. [^2]: Mokyr's work spans economic history and the history of technology. His books "The Lever of Riches" and "A Culture of Growth" trace how European institutions enabled the Industrial Revolution. [^3]: Previous cross-country comparisons of working hours relied on inconsistent definitions and incomplete coverage. The new database standardizes measurement across countries and time. [^4]: The capital asset pricing model (CAPM) and related theories prescribe discount rates based on systematic risk. Empirical discount rates often exceed theoretical predictions by several percentage points, suggesting behavioral or organizational explanations. [^5]: Hulten's theorem states that the aggregate impact of sector-specific productivity gains equals the productivity improvement times the sector's share in GDP. This limits the macroeconomic impact of even dramatic improvements in narrow applications. [^6]: Forecast evaluation requires appropriate benchmarks. Random walk forecasts, simple autoregressive models, and survey expectations provide reference points for assessing central bank performance. [^7]: DICE (Dynamic Integrated Climate-Economy) is an integrated assessment model that links economic growth, emissions, climate change, and damages in a unified framework for cost-benefit analysis of climate policy. [^8]: The World Inequality Lab, founded by Thomas Piketty and colleagues, maintains the World Inequality Database (WID), which provides consistent series on income and wealth distribution across countries and time. [^9]: "Crosstagion" describes bidirectional contagion between traditional and decentralized finance through shared participants, correlated assets, and interconnected protocols. [^10]: CBDCs are digital liabilities of central banks, distinct from commercial bank deposits and cryptocurrencies. They aim to provide the safety of central bank money with the convenience of digital payments. [^11]: Geoeconomic fragmentation refers to policy-driven reversal of global economic integration, distinct from market-driven changes in trade patterns. It involves deliberate use of economic tools for geopolitical objectives.