VoxTalks Economics

On 2 April 2025, the United States imposed tariffs on almost every country on earth. The next day, China responded with export controls on the entire world. In the space of one week, world trade had been weaponised as it has never been in peacetime.
Richard Baldwin of IMD Business School, the founder of VoxEU and a former president of the Centre for Economic Policy Research, wrote World War Trade to make sense of the events of the last 12 months. The dramatic April salvos have settled into a trade Cold War; US tariffs and Chinese export controls are lodged in place, with neither side expecting the other to back down. 
And yet world trade grew in 2025; exports from every country rose except from the US, which recorded its largest trade deficit. The rest of the world is self-organising a new order. When one country joins a rules-based regional agreement, the cost of staying out rises for the next. EU-Mercosur and EU-Australia deals, stalled for years, crossed the line. An expanding CPTPP and early alignment talks between the EU and CPTPP blocs are pulling more partners in. The old system was a cathedral built and maintained largely by the US; the architect burned it down. Something else is being built in its place.
The book discussed in this episode:
Baldwin, Richard. 2026. World War Trade: Conflict, Containment, and the Emergent World Trading Order. Rapid Response Economics 6. CEPR Press. Free to download from CEPR Press.
To cite this episode:
Phillips, Tim, and Richard Baldwin. 2026. "World War Trade." VoxTalks Economics (podcast). 

Assign this as extra listening. The citation above is formatted and ready for a reading list or VLE.

About the guest

Richard Baldwin is Professor of International Economics at IMD Business School in Lausanne. He founded VoxEU, the Centre for Economic Policy Research's policy portal, and served as president of CEPR. His research spans trade policy, globalisation, and the political economy of trade; he is one of the architects of modern thinking on global value chains and the "second unbundling" of production. World War Trade is the sixth book in the CEPR Press Rapid Response Economics series.
Research cited in this episode

TACO (Trump Always Chickens Out) began as a joke in finance markets as a description of the pattern in which the US president announces aggressive trade measures and then partially or fully reverses them when markets react or negotiations begin. Baldwin argues that financial markets eventually priced in a TACO floor; once they believed Trump would back down before a full market meltdown, they stopped reacting to his escalations as if they were terminal. The dynamic makes tariff threats simultaneously more frequent and less credible.
Domino regionalism describes the self-reinforcing logic by which regional trade agreements attract new members. When one economy gains preferential access to a large market, the cost of staying outside that agreement rises for its trading partners; that pressure brings in the next country, which raises the cost for the next, and so on. Baldwin identified this mechanism in the regional trade wave of the 1990s and argues it is now operating again, accelerated by the uncertainty created by US and Chinese trade weapons. The EU-Mercosur deal unblocking was the trigger; EU-Australia followed within weeks.
G-0 world is a concept developed by political scientist Ian Bremmer to describe a world in which no single country or group of countries provides consistent global leadership. Baldwin draws on this framework to explain why regional conflicts and trade disputes have become harder to contain since the US began stepping back from its hegemonic role; the trade cold war is one expression of that leadership vacuum, but so is the reduced capacity to broker deals in the Middle East or manage the Black Sea grain corridor.
CPTPP (Comprehensive and Progressive Agreement for Trans-Pacific Partnership) is a rules-based regional trade agreement covering eleven countries across Asia and the Pacific, including Japan, Canada, Australia, Vietnam, and the United Kingdom. It operates without US or Chinese membership and maintains deep disciplines on intellectual property, investment, and trade in services. Baldwin identifies it, alongside the EU, as one of the two main "pools of predictability" around which the new post-war trading order is forming. The two blocs have opened alignment discussions that, if concluded, would bring a very large share of world trade under compatible rules.
RCEP (Regional Comprehensive Economic Partnership) is a large but shallower regional agreement covering much of Asia, including China, Japan, South Korea, Australia, and the ten ASEAN nations. It involves Chinese leadership and does not carry the depth of disciplines found in CPTPP. Baldwin notes that it is rules-based and that as long as China plays by those rules it could enlarge; but it has not attracted the same wave of new joiners as CPTPP and the EU framework.
The EU Anti-Coercion Instrument is a European Union mechanism, adopted in 2023, allowing the EU to retaliate against third countries that use trade or economic measures to coerce member states into changing their policies. Baldwin cites it as an example of the "building bunkers" response adopted by many economies; rather than retaliating directly against US tariffs, countries are changing their domestic laws to give themselves tools to counter future coercion without breaching WTO rules.
More VoxTalks Economics episodes

This is the second time Richard Baldwin has discussed the 2025 trade upheaval on VoxTalks Economics. He appeared alongside Gene Grossman of Princeton in What's Next for Trump's Tariffs, broadcast in January 2026, which covered the seismic moves of 2025 as they were unfolding.

"When you look at the world now, does it look more uncertain or less uncertain?" In December 2025, the Bank of England's Financial Policy Committee (FPC) answered that question by cutting the equity capital requirement for UK banks. David Aikman (NIESR) and John Vickers (University of Oxford), two former senior Bank insiders who helped to design the regulatory framework post-GFC, think the committee got it wrong.
The FPC lowered the benchmark capital requirement from 14% to 13% of risk-weighted assets, a move that could free up roughly £30 billion of capital across the UK banking system. Aikman and Vickers see no compelling economic reason for the change. They argue that the 2015 benchmark was already set too low, built on questionable assumptions about how well resolution frameworks would work. Since 2015, Brexit, the pandemic, and a sharply stretched fiscal position have all increased the likely cost of a future crisis. The practical effect of the loosening may not even be more lending, but higher dividends and share buybacks. And the December decision may signal a weakening of the leverage ratio backstop, the constraint that limits bank borrowing regardless of how risk weights are applied.
The research behind this episode:
Aikman, David, and John Vickers. 2026. "The Bank of England's Capital Mistake." VoxEU, 15 January 2026. 
To cite this episode:
Phillips, Tim, David Aikman, and John Vickers. 2026. "The Bank of England's Capital Mistake." VoxTalks Economics (podcast). 

Assign this as extra listening. The citation above is formatted and ready for a reading list or VLE.

About the guests

David Aikman is Director of the National Institute of Economic and Social Research (NIESR). He worked at the Bank of England from 2003 to 2020, where he served as Technical Head of Division in Financial Stability and was centrally involved in the creation of the Financial Policy Committee. His research spanning macroprudential regulation, systemic risk, and the macroeconomics of financial crises has made him one of the leading academic voices on bank capital policy in the UK.
Sir John Vickers is Warden of All Souls College and Professor of Economics at the University of Oxford. He served as Chief Economist and a member of the Monetary Policy Committee at the Bank of England, and chaired the Independent Commission on Banking from 2010 to 2011, which recommended substantially higher capital requirements than those subsequently adopted. His research spanning industrial economics, competition policy, and financial regulation has shaped UK banking policy for two decades.
Research cited in this episode

Equity capital requirements specify the minimum proportion of a bank's assets that must be funded by shareholders' equity rather than borrowed money. Equity is the only form of funding that can absorb losses without triggering insolvency: if a bank suffers unexpected losses, its shareholders bear them first. In the run-up to the 2008 financial crisis, some large institutions held equity equivalent to as little as two or three percent of their total exposures, implying leverage of up to forty times; a small shock was enough to render them insolvent. The post-crisis repair effort was designed to ensure that could not happen again.
Risk-weighted assets (RWAs) are the denominator against which capital requirements are measured. Rather than applying the capital ratio to the raw value of all assets, the framework deflates each asset by an estimated risk factor: a mortgage backed by collateral is treated as less risky than an unsecured corporate loan, for example. Capital requirements are then expressed as a percentage of this risk-adjusted total. The approach creates significant complexity and depends heavily on the accuracy of the risk weights; much of the story of 2008 was that regulators allowed banks to attach implausibly low risk weights to their exposures, understating the true leverage in the system.
The Financial Policy Committee (FPC) is the Bank of England body responsible for macroprudential oversight of the UK financial system. Created in 2013, it sits above the individual regulators to take a system-wide view of whether risks are building and whether the financial system as a whole has adequate resilience. One of its primary tools is setting the overall capital requirement benchmark for UK banks. In 2015 it set that benchmark at 14% of risk-weighted assets; in December 2025 it reduced it to 13%.
The leverage ratio is an alternative measure of bank capitalisation that does not apply risk weights. It expresses equity as a simple percentage of total assets, regardless of what those assets are. The UK leverage ratio backstop currently stands at around 3 to 4%, implying maximum leverage of roughly twenty-five to thirty times for systemically important banks. Vickers and Aikman note that for some UK banks the backstop has become the binding constraint, which they regard as a warning sign: it suggests that risk-weighted measures are understating actual leverage, not that the backstop should be relaxed.
Resolution frameworks are the legal and operational mechanisms that allow regulators to manage the failure of a bank without a taxpayer bailout, by imposing losses on shareholders and creditors in an orderly way. A central assumption in the FPC's 2015 capital benchmark was that resolution would work effectively in a future crisis, which justified a lower capital requirement. Vickers and Aikman are sceptical: the experience of Credit Suisse in 2023, which required a state-assisted rescue despite the existence of resolution plans, illustrates that orderly resolution of a major institution cannot be taken for granted.
Basel 3.1 is the latest package of international banking regulatory standards agreed by the Basel Committee on Banking Supervision, designed to address weaknesses in how risk weights are calculated. Its implementation in the UK is scheduled for 2027, nineteen years after the 2008 crisis. The FPC's December 2025 decision is partly contingent on Basel 3.1 being implemented as planned; Aikman notes that there have been repeated international delays and rollbacks, and that the UK's ability to move ahead unilaterally is constrained by what other major jurisdictions do.
The 2023 banking stress saw three US regional banks (Silicon Valley Bank, Signature Bank, and First Republic) fail in quick succession in March 2023, followed by the forced rescue of Credit Suisse by UBS. These events occurred in what was, by historical standards, a relatively stable macroeconomic environment. Vickers cites them as evidence that banking sector vulnerabilities have not been eliminated by post-2008 reforms, and as a caution against complacency about the effectiveness of current safeguards.
More VoxTalks Economics

Making banking safe Our financial system is supposed to be more resilient than before the global financial crisis, but that didn’t save Silicon Valley Bank, Signature Bank or First Republic. So what went wrong, and can we fix it? Steve Cecchetti and Kim Schoenholtz suggest how regulators can make banking safer.

Two explanations circulated immediately after the March to Save America on January 6, 2021 turned into a riot: a mob manipulated by a demagogue, or ordinary citizens defending democracy against a stolen election. Konstantin Sonin, David Van Dijcke, and Austin Wright have used anonymised location data from forty million mobile devices to investigate why the protests escalated so dramatically.
No surprise: partisanship was the strongest predictor of attendance, proximity to Proud Boys chapters and use of the far-right social network Parler also increased participation. But political isolation amplified the movement: the communities most over-represented among those who traveled to Washington were small Republican enclaves surrounded by Democrat-leaning areas, politically and socially cut off from their neighbours. And participation also spiked in counties that experienced a "midnight swing," where the reported vote count favoured Trump on election night before shifting to Biden as mail-in ballots were counted. These were precisely the counties where the "Stop the Steal" narrative landed hardest. 
The research behind this episode:
Sonin, Konstantin, David Van Dijcke, and Austin L. Wright. 2023. "Isolation and Insurrection: How Partisanship and Political Geography Fueled January 6, 2021." CEPR DP18209. 
To cite this episode:
Phillips, Tim, and Konstantin Sonin. 2026. “What triggered January 6?” VoxTalks Economics (podcast). 

Assign this as extra listening. The citation above is formatted and ready for a reading list or VLE.

About the guest

Konstantin Sonin is the John Dewey Distinguished Service Professor at the Harris School of Public Policy at the University of Chicago. Born in the Soviet Union, he has spent his career studying how political institutions work under stress, with particular attention to how information and misinformation shape political behaviour, elections, and collective action. He is one of the leading economists working on the political economy of authoritarian and democratic governance, and his research on protest, polarisation, and political geography has made him a central figure in the study of democratic backsliding.
Research cited in this episode

Regression discontinuity design is a statistical method used to identify causal effects by exploiting a threshold or cutoff. Sonin, Van Dijcke, and Wright use two regression discontinuity designs: one exploiting the narrow margins by which Trump lost certain states, and one exploiting the gap between the election-night vote tally and the final certified result in individual counties. In both cases, the design allows them to isolate the effect of a specific trigger on protest participation, separating it from the general background of partisan feeling.
The "midnight swing" refers to the shift in reported vote tallies that occurred in many counties on election night 2020 as large batches of mail-in ballots were counted. Because mail-in voters skewed heavily Democratic, counties where in-person votes were reported first showed strong Trump leads that reversed overnight as the mail-in totals arrived. For professional observers and election administrators, this pattern was entirely expected; it followed directly from the different rules different states used to count mail-in ballots during the pandemic. For many voters, particularly those already primed to distrust the electoral process, it read as suspicious. The paper finds that communities exposed to larger swings sent disproportionately more participants to Washington on January 6.
Network Exposure design is a methodological innovation introduced in this paper. It measures how much exposure a given community had to election-denial signals flowing through its social networks, and distinguishes this from exposure arising simply through geographic proximity to other communities. Isolated communities proved hypersensitive to information traveling through their social networks, but not to information spreading through neighbouring areas. This suggests the amplification mechanism was social, not spatial.
Political isolation in this paper refers to being a minority political community within a larger, differently-leaning area. A small Republican-voting enclave inside a Democrat-leaning county or district is politically isolated in this sense. The paper finds that isolation of this kind was a strong amplifier of partisanship in predicting participation. Two other measures of isolation, one based on mobile device travel patterns ("locational isolation") and one based on Facebook connections ("social media isolation"), produce consistent results, suggesting the effect is not an artefact of how isolation is measured.
The Proud Boys are a far-right extremist organisation active in the United States. The paper finds that communities with a local Proud Boys chapter were over-represented among those who traveled to Washington on January 6, making proximity to the organisation a robust correlate of participation, independent of general partisan leanings.
Parler was a social media platform popular among far-right users in the United States during the period leading up to January 6, 2021. Communities where Parler usage was relatively higher were also over-represented among participants in the March to Save America, suggesting that the platform played a role in amplifying mobilisation signals within the networks most susceptible to them.
Collective action theory is the study of how individuals decide to participate in group action, particularly when the costs fall on participants individually but the benefits are shared. Sonin, Van Dijcke, and Wright contribute behavioural evidence on the specific role of political isolation and network-amplified grievance in driving participation.
More VoxTalks Economics

The Grievance Doctrine What if trade policy wasn’t really about trade at all? What if it was about revenge, power, and punishment, tariffs as tantrums and diplomacy as drama? Richard Baldwin on what is driving the US policy agenda. 
How protests are born, and how they die Every year we see thousands of protest movements on our city streets. Benoît Schmutz-Bloch explains why do some protests persist, and some disappear, and some remain peaceful, but others become violent.

Every Bitcoin transaction needs to be verified on the blockchain. There is no central authority that does this, but Bitcoin's blockchain has run uninterrupted since 2009 and now carries a market capitalisation of $1.3 trillion, roughly 4% of US GDP. Its original promise was more radical: that we do not need a trusted intermediary to spend money, write contracts, or create finance. In the fifth LTI report, published today, Yackolley Amoussou-Guenou, Bruno Biais, and Sara Tucci-Piergiovanni ask how much of that promise has held. 
Bruno talks to Tim Phillips about blockchain’s potential, its flaws, and its future.  
It is a Nash equilibrium: if you believe others will follow the rules, it is in your interest to follow them too. On that foundation Bitcoin’s ledger has been running continuously for 16 years. Smart contracts, pioneered by Vitalik Buterin's Ethereum, extend the logic to financial agreements. Decentralised finance promised to cut out rent-seeking intermediaries. Cryptocurrencies can step in where banks are broken or currencies have collapsed; in Lebanon, when bank accounts were frozen and payments stopped, businesses switched to crypto and kept operating. 
But the technology's libertarian origins may need to be sacrificed: As Bruno says, without transparency there is no trust, and transparency in this market may require regulation.
The research behind this episode:
Amoussou-Guenou, Yackolley, Bruno Biais, and Sara Tucci-Piergiovanni. 2026. "Can Blockchain Decentralize Money, Contracts, and Finance?" LTI Report 5. CEPR and Long-Term Investors@UniTo. Freely available to download at cepr.org. 
To cite this episode:
Phillips, Tim, and Bruno Biais. 2025. "Can Blockchain Decentralize Money, Contracts, and Finance?" VoxTalks Economics (podcast). 

Assign this as extra listening. The citation above is formatted and ready for a reading list or VLE.

About the guest

Bruno Biais is Professor of Finance at HEC Paris and a Research Fellow at the Centre for Economic Policy Research (CEPR). His research spanning financial market microstructure, corporate finance, and the economics of blockchain has made him one of the leading economists working at the intersection of finance and decentralised technology. He has studied blockchain and cryptocurrency markets since their early years, and his theoretical models of consensus mechanisms and cryptocurrency valuation have shaped how economists understand the conditions under which decentralised systems can and cannot sustain themselves.
Research cited in this episode

The blockchain is a distributed ledger maintained by a network of nodes, each holding an identical copy of the record of ownership. When a transaction is submitted, all nodes verify it against the existing ledger and update their copies to reach consensus on the new state. No central authority manages this process; its stability rests entirely on the incentive structure built into the protocol.
Nash equilibrium is a concept from game theory, named for the mathematician John Nash, describing a situation in which each participant's strategy is the best response to the strategies of all others; no individual has an incentive to deviate unilaterally. Biais and co-authors identify the Bitcoin protocol as a Nash equilibrium: if you believe others will follow the rules, it is in your own interest to follow them too. That self-reinforcing alignment of incentives, rather than goodwill or central enforcement, is why the blockchain has remained valid since 2009.
Smart contracts are lines of code deposited on a blockchain that execute automatically when specified conditions are met: if X, then Y. Vitalik Buterin introduced them through the Ethereum platform, which offers a richer programming language than Bitcoin and allows users to hold collateral on-chain to guarantee the contract will pay out. Smart contracts underpin automated market makers, decentralised lending, and a wide range of financial applications that require no counterparty or intermediary to enforce the agreement.
Oracles are third-party services that transmit data about real-world events to a blockchain, allowing smart contracts to respond to things that happen off-chain. A contract that pays out when a house burns, for example, requires an oracle to report that event to the network. Oracles introduce a point of fragility: the authenticity and accuracy of off-chain information must be established before the network accepts it, and that verification is more vulnerable to error and manipulation than the on-chain consensus mechanism itself.
Front-running and miner extractable value (MEV) describe the practice by which technically sophisticated actors exploit the public visibility of pending transactions to extract profits at the expense of ordinary users. Because transactions on public blockchains are broadcast to all nodes before they are confirmed, an actor who sees a large pending purchase can execute the same trade first, drive the price up, and then sell at a profit once the original transaction goes through. The cost falls on the smaller trader. Biais notes that the barriers to entry and economies of scale in this activity have concentrated power in the hands of a small, technically skilled group, recreating the kind of intermediary rents that decentralised finance was designed to eliminate.
Automated market makers are smart contracts that provide continuous liquidity for trading between two assets by holding reserves of both in a pool and setting prices according to the ratio of the reserves. A large purchase of one asset depletes that side of the pool and raises its price; a large sale depresses it. Automated market makers have become a central mechanism of decentralised finance, replacing the order-book systems used in traditional exchanges.
Stablecoins are cryptocurrency tokens designed to maintain a fixed value relative to a conventional currency, typically the US dollar. They are issued by private entities that hold reserves intended to back the peg. Tether, the largest stablecoin by market capitalisation, holds its reserves in a mix of Treasury bills, Bitcoin, and precious metals; in 2021, the US Commodity Futures Trading Commission fined Tether for misrepresenting those reserves and required it to disclose their composition, making this information publicly available for the first time. Dai is an algorithmically managed stablecoin that maintains its peg through over-collateralisation in cryptocurrency rather than conventional reserves.
The Diamond-Dybvig model is a theoretical framework developed by Douglas Diamond and Philip Dybvig explaining why financial intermediaries that hold illiquid assets while issuing liquid claims are inherently vulnerable to runs. When enough depositors demand withdrawal simultaneously, the institution is forced to sell assets at a loss, making further withdrawals impossible and confirming the fears that triggered the run. Biais applies this logic to stablecoins: if enough holders attempt to redeem simultaneously, the issuer must sell its reserves in volume, driving down their price and potentially breaking the peg.
Central bank digital currencies (CBDCs) are digital tokens issued and managed by central banks, distinct from both commercial bank deposits and private stablecoins. Biais distinguishes two potential use cases: retail CBDCs, which would allow individuals to hold central bank money directly, and wholesale CBDCs, which would facilitate settlement between large financial institutions. He regards the wholesale application as the more promising; a wholesale CBDC could enable fast, low-cost atomic settlement of cross-currency transactions between banks under central bank oversight, a significant improvement on current interbank settlement systems.
MiCA (Markets in Crypto-Assets Regulation) is the European Union's regulatory framework for crypto-asset service providers, which came fully into force in December 2024. It requires licensing for issuers and service providers operating within the EU and imposes disclosure, reserve, and conduct requirements intended to align the sector more closely with the standards applied in traditional financial markets.
Hayek's currency competition refers to the argument by Friedrich Hayek that competition between privately issued currencies would discipline monetary policy: users would switch away from currencies managed irresponsibly, and that threat would encourage better central bank behaviour. Biais applies this argument to cryptocurrencies and stablecoins in countries where the domestic currency has been mismanaged. He cites Nigeria, where sharp depreciation of the naira was accompanied by rising crypto adoption; over the following period, Nigeria's central bank raised interest rates and created a more transparent foreign exchange market. Biais suggests, tentatively, that the competitive pressure from crypto alternatives may have contributed to that improvement.
More VoxTalks Economics

Do stablecoins threaten financial stability? Stablecoins are digital tokens, pegged to a fiat currency. What could possibly go wrong? For one type of stablecoin the answer is: plenty, according to Richard Portes. 
In coin we trust Crypto investors make a lot of noise, but who are they, and do they behave differently to other retail investors?
Do cryptocurrencies matter? Can cryptocurrencies be useful? Not just for crypto bro speculators, but as a shield against the depreciation of the official currency if a government is determined to pursue inflationary policies.

Could AI transform our economies to produce explosive growth? Most economists are sceptical at best. Anton Korinek of the University of Virginia, leader of the CEPR research policy network on AI, thinks the threshold is closer than those models suggest.
In his latest work, Korinek, Tom Davidson, Basil Halperin, and Thomas Houlden, have built a growth model that captures what happens when AI starts automating AI research itself. Automation does two things simultaneously: it accelerates research, and it offsets the diminishing returns that have historically stopped self-improving processes from compounding. Three reinforcing feedback loops: software quality, hardware quality, and general technological progress, each amplify the others. Korinek's findings are more optimistic than even the AI labs' own roadmaps, which focus on software capability alone. 
The research behind this episode:
Davidson, Tom, Basil Halperin, Thomas Houlden, and Anton Korinek. 2026. "When Does Automating AI Research Produce Explosive Growth? Feedback Loops in Innovation Networks." Working paper, January 2026.
To cite this episode:
Phillips, Tim, and Anton Korinek. 2026. "When Does Automating AI Research Produce Explosive Growth?" VoxTalks Economics (podcast). 

Assign this as extra listening. The citation above is formatted and ready for a reading list or VLE.

About the guests

Anton Korinek is a professor of economics at the University of Virginia. He leads the CEPR Research Policy Network on AI, which is building a community of researchers to understand and anticipate the economic impact of artificial intelligence. He is a member of Anthropic's Economic Advisory Council and was named by Time magazine among the hundred most influential people in AI. His research spanning the economics of transformative AI, growth theory, and the implications of advanced automation for labor markets and inequality has made him one of the most widely cited economists working on these questions. He is also the founder of the Economics of Transformative AI initiative at the University of Virginia, which focuses on the long-run economic consequences of AI systems that approach or exceed human-level capabilities.
Visit the CEPR Research Policy Network on AI.
Research cited in this episode

Daron Acemoglu's estimate of AI's growth impact. Acemoglu calculated that AI would raise annual growth by approximately 0.07 percentage points, arriving at this figure by multiplying the share of jobs likely to be affected by AI, the fraction of tasks within those jobs that AI could perform, and the productivity gain per task. Korinek argues the estimate was a reasonable description of the AI that existed in 2024 but did not account for the trajectory of capabilities since, nor for the feedback loops between AI progress and further AI development that his own paper models.
Recursive self-improvement. The idea that an AI system, once capable enough, could design improved versions of itself, triggering an accelerating cycle of capability gains. The concept was first articulated by John von Neumann in the 1950s and has since become central to debates about transformative AI. All major AI labs, Korinek notes, are working towards some version of this vision; the economic question is whether the resulting growth would be explosive or would be damped by diminishing returns.
Semi-endogenous growth models. A class of economic growth models in which long-run growth depends on the scale of the research workforce and the returns to research effort. The canonical insight, associated most closely with Nicholas Bloom and co-authors, is that "ideas get harder to find"; maintaining a given rate of progress requires ever-increasing research investment. Korinek and co-authors use and extend this framework, showing that automation can counteract diminishing returns by replacing human labor with capital in the research process, creating a new feedback loop that was absent from earlier models.
Kaldor's balanced growth facts. Nicholas Kaldor's observation, made in the mid-twentieth century, that the major macroeconomic aggregates, including the capital-output ratio, the labor share of income, and the rate of return to capital, remain roughly stable over long periods. Growth economists built their models, including the Solow and Ramsey models, to fit these regularities. Korinek notes that those models were appropriate precisely because they matched the historical data; the question his paper raises is whether the data of the next few decades will look different enough to require a different class of models.
Moore's Law. The empirical regularity, observed in computing hardware since the 1960s, that the number of transistors on a chip approximately doubles every two years. Korinek uses chip progress as a calibration benchmark: maintaining that rate of doubling has historically required roughly an eight percent annual increase in the scientific workforce working on chips. This figure allows the model to be parameterised with a real-world measurement of how much additional research input is needed to sustain a given rate of technological progress.
Consumer surplus from digital technologies. Korinek raises the problem that GDP statistics are designed to measure market transactions and therefore do not capture the value people derive from digital goods and services beyond what they pay for them. He references research from the Stanford Digital Economy Lab as an example of work attempting to quantify this surplus. The implication for the paper's argument is that explosive AI-driven growth could be underestimated even in the statistics used to monitor it.
More VoxTalks Economics episodes

"Our Workless Future", an earlier conversation with Anton Korinek from September 2022, in which he set out the case for taking AI's impact on labor markets seriously.
Related reading on VoxEU

Firms predict an AI productivity boom is coming, a survey of over 5,000 CFOs, CEOs, and executives shows that around 70% of firms actively use AI, particularly younger, more productive firms. They forecast AI will boost productivity by 1.4%, increase output by 0.8%, and cut employment by 0.7% over the next three years.
How AI is affecting productivity and jobs in Europe, firm-level evidence on AI’s effects in Europe. The authors find that AI adoption increases labour productivity levels by 4% on average in the EU, with no evidence of reduced employment in the short run.
From AI investment to GDP growth: An ecosystem view, how the current AI wave is contributing to US GDP, both directly through investment and indirectly through ongoing service flows.