What Makes Economies Grow

This is a typological theme, organising the development economists of Conversations with Tyler — Daron Acemoglu, Brad DeLong, Yasheng Huang, Dan Wang, Joe Studwell, Dani Rodrik, Jason Furman, Abhijit Banerjee, and Raj Chetty — along the oldest question in the discipline: what makes an economy grow, converge on the rich-world frontier, and lift its people — and what leaves another stalled for decades. The speakers broadly agree on what does not explain it; they divide sharply on what does. Synthesised from nine episodes of Conversations with Tyler.

The organising axis

Start with the negative consensus, because it is unusually strong. Geography is not destiny: the correlation between latitude and income is, for Daron Acemoglu, an artefact of which territories Europeans colonised and how, and the Reversal of Fortune — places rich in 1500 are poor today, and the reverse — is a decisive test against any purely geographic theory, since geography did not change while relative prosperity did. There is no transferable recipe: Dani Rodrik ‘hates’ best-practice institutions and quotes his teacher Avinash Dixit — ‘the world is second best, at best’ — while Abhijit Banerjee dismisses Singapore, Dubai, and Ireland as tiny, idiosyncratic, politically lucky cases, not templates. And convergence is getting harder, not easier: the rapid East Asian path that ran Japan, South Korea, Taiwan, and China to wealth is, most of the cluster agrees, narrowing or closed.

The agreement ends there. Pressed on the binding constraint — the one variable that, relaxed, would let growth run — the cluster splits three ways.

The institutions camp holds that what decides an economy’s fate is the rules of the game: whether power is contested, whether the discovery and deployment of useful knowledge is rationalised, whether the most capable people are free to challenge authority or captured by it. The structural-transformation camp holds that growth is a mechanism — moving labour out of subsistence and into higher-productivity, tradable activity, historically manufacturing — and that industrial policy is the proven engine. The people-and-place camp holds that the constraint is the human input itself: demographic density, human capital, confidence at the very bottom, and the cross-class contact that converts a child’s talent into a life.

Cutting across all three is a methodological fault line that the cluster argues about openly: can growth be engineered at all, or only watched, midwifed, and explained after the fact?

Position one: institutions are the master variable

Daron Acemoglu, Brad DeLong, Yasheng Huang, and Dan Wang each locate the decisive variable not in resources or geography but in the rules that govern power, knowledge, and talent.

Acemoglu’s frame is the narrow corridor. Most of history falls into one of two failure modes — stateless societies with no public goods, or despotic states with capacity but no societal check. Liberty, and the prosperity that travels with it, lives in a fragile corridor where state capacity and social mobilisation grow together in competitive tension, the Red Queen effect in which each side runs to stay in place. The European version traces to a collision: Germanic norms of assembly politics, in which chiefs answered to the governed, fused with the surviving Roman administrative machinery — Byzantium had the bureaucracy and the Christianity but not the bottom-up element, and produced no comparable liberty. The mechanism by which institutions pay off is concrete and fiscal: countries that democratise add roughly 20–25 per cent to GDP per capita over the following decades, mainly by taxing and spending more on education and health. See Daron Acemoglu on the Struggle Between State and Society.

DeLong supplies the long-run version. The ‘long twentieth century’ (1870–2010) is humanity’s one era of sustained 2 per cent growth, and what made it was the rationalisation of discovery: three institutions — the industrial research lab, the modern corporation, and globalisation — that earlier waves of progress, dependent on lone tinkerers, had lacked. German universities trained chemists to manipulate molecules while Britons learned to write verse in the style of Macaulay; large intermediaries like Deutsche Bank and the Morgan partnership made a routine business of ‘intermediating information between people who had money and people who had ideas’. Behind even this sits a deeper institutional accident — the Royal Society’s decision to settle questions through experiment alone, displacing the older world in which an idea’s usefulness to the ruling machine mattered more than its truth. Progress, in DeLong’s telling, comes from finding ‘much better institutions in fairly discrete steps’, not from population size. See Brad DeLong on Intellectual and Technical Progress.

Yasheng Huang and Dan Wang converge on the same lever from the Chinese case, and identify the same culprit: the institution that decided who the smartest people became. Huang’s argument in The Rise and Fall of the EAST is that the civil service examination — meritocratic in execution, with double-anonymised marking — was destructive precisely where it worked, because it homogenised the intellectual class into a frozen, backward-looking Confucianism and absorbed every ambitious mind into the system rather than against it. The Chinese state was simultaneously overdeveloped (the only institution, with no independent church, intelligentsia, or merchant class beside it) and underdeveloped (never able to tax broadly or sustain public services). His tell is the period he most admires: the fragmented interregnum of 220–581 AD, with no ideological monopoly and competing states, which posted China’s historical peak in inventions per capita. See Yasheng Huang on the Development of the Chinese State. Wang reaches the identical diagnosis from the politics of liberalism: China never produced a lawyerly tradition because no institution — no pope, no landed aristocracy, no guild — could constrain the court, so the exam funnelled all energy toward loyalty to the centre, and no one advanced by arguing for limits on imperial power. See Dan Wang on What China and America Can Learn from Each Other.

Position two: structural transformation is the engine

Joe Studwell, Dani Rodrik, and — for the frontier rather than the catch-up case — Jason Furman hold that growth is less a matter of rules than of mechanism: shifting labour into higher-productivity, tradable work, and that this can be deliberately engineered.

Studwell’s is the most confident statement. Industrial policy — cheap money aimed at manufacturing, temporary protection, reward for exporters, and crucially competition — has ‘been refined well over a century’ and is the only mechanism that reliably raises an economy’s technological level to the point of convergence. India failed not because the tool is bad but because Mahalanobis-era planners picked winners and never subjected them to competition; Brazil applied it patchily. He is unworried that robots foreclose the path for Africa: factory labour at $60–65 a month against China’s ~$600, plus a garment robot’s six-figure upfront cost and inability to flex with demand, keeps the door open. His one heterodox addition is demographic — Africa’s binding constraint was never mainly governance but population density, a fifth of Asia’s in 1960, because you cannot mobilise people who are not there. See Joe Studwell on Africa, Asia, and What Development Actually Requires.

Rodrik gives the microfoundations and the warning. Manufacturing is special for three concrete reasons: its technology is portable across borders in a way agriculture and services are not; it absorbs unskilled labour straight off the farm (early footwear plants needed eye–hand coordination, not literacy) and trebles a worker’s productivity; and it is tradable, so one segment can master world markets without waiting for the whole economy to rise. But automation and Chinese competition have triggered premature deindustrialisation — poor countries shedding factory jobs at far lower incomes than the rich world ever did — which closes the one engine that delivered 4–5 per cent per-capita growth for decades. His response is to rename and re-aim the tool: ‘structural transformation policy’, oriented to moving up the economy generally including into tradable services, and measured by employment rather than output, exports, or patents. See Dani Rodrik on Premature Deindustrialization and Why the World is Second Best, at Best.

Furman applies the same logic to a frontier economy that has already transformed: the task is keeping the growth engine dynamic. His diagnosis of America’s productivity slowdown is that rising market concentration suppresses business investment — monopoly rents remove both the pressure and the need to invest — and that ‘merger to monopoly’ (Facebook, Google) deserves more scrutiny than organic scale (Walmart, Amazon). His remedies are talent and fluidity: asked to name priorities as a hypothetical productivity czar, he answers ‘more immigration’ three times before anything else, then federal R&D, immediate expensing, and YIMBY housing reform so workers can actually move to high-productivity cities. The China-shock literature, he argues, teaches us about labour-market rigidity — occupational licensing, employer-tied healthcare, land-use constraints — not about trade. See Jason Furman on Productivity, Competition, and Growth.

Keyu Jin supplies a concrete mechanism for this camp from inside the Chinese case: the Mayor Economy, in which political centralisation and economic decentralisation combine so that local officials compete against each other on a yardstick — GDP growth, then environmental compliance, then technological innovation — set and enforced by Beijing through the power to promote, demote, or dismiss them. Where Studwell describes industrial policy at the level of national strategy, Jin’s account supplies the local-official incentive structure that actually executes it, including the same acknowledged cost the camp accepts elsewhere — wasted capital and misallocation while political rather than market judgement is picking winners. See Keyu Jin on China's Economy, Industrial Policy, and State Capitalism.

Position three: the constraint is people and place

Raj Chetty, Abhijit Banerjee, and Studwell’s demographic argument locate the binding constraint in the human input — talent, confidence, density, and the social geography that determines whether ability is used or wasted.

Chetty’s work moves the question inside a rich country and finds growth that does not lift its people. Absolute mobility — the chance of earning more than your parents — has fallen from roughly 90 per cent for the 1940 birth cohort to about 50 per cent for the 1980s cohort, and the driver is rising inequality, not slow growth: the same growth distributed as in 1970 would restore most of the gap. Where mobility is high, the strongest correlate is low segregation — actual cross-class contact, which a small Iowa town forces and a large city lets the affluent escape. And the mechanism by which opportunity transmits is subtle: a high-value-added kindergarten teacher raises adult earnings not through academic content, which fades by third grade, but through non-cognitive skills the labour market prices and tests do not. See Raj Chetty on Teachers, Social Mobility, and How to Find Answers to Big Questions.

Banerjee turns to the very bottom and finds the constraint is confidence. In the multi-country graduation programme, a cash grant plus training and coaching produced returns of several hundred per cent, but the cash-only arm failed in Ghana — the coaching was load-bearing. For people who have never succeeded at anything and are treated with contempt, the intervention works by converting a daunting, vague task into a concrete sequence of steps (‘here is how you get on a bus, here is the wholesale market’) and by signalling that someone believes they can do it. His structural diagnosis is bleaker: the real development problem is the petty entrepreneur, the subsistence micro-business that UBI and microcredit make marginally bigger, when the historical mechanism that worked was the opposite — letting a few firms grow large enough to employ the rest — and nobody has a current model for that transition. See Abhijit Banerjee on Theory, Practice, and India.

Studwell’s population-density argument belongs here too: people are the key input in a poor country with no cash and no technology, and Africa’s optimism today is private and agricultural precisely because it needs least of the weak state — farmer-led irrigation, not state megaprojects, has driven the world’s fastest agricultural growth since 2000.

The cross-cutting axis: can growth be engineered?

The cluster’s sharpest live disagreement is methodological, and it does not respect the three camps. Joe Studwell thinks growth can be manufactured — industrial policy is a refined, repeatable tool, and the failures are failures of execution. At the opposite pole, Abhijit Banerjee says flatly that economists do not understand growth: the one durable finding is Robert Solow’s — fast growth runs out as a country exhausts its best talent, capital, and locations — but why Brazil grew for two decades and then stalled for two more is genuinely not known, and confident recipes are a delusion.

Between them sits a debate about how we know anything. Banerjee defends the randomised controlled trial, but not for the reason usually given: its value is not unbiased estimation (‘a red herring’) but that a theorist can keep redesigning the treatment to interrogate the hypothesis in his head. Dani Rodrik judges RCTs overrated and crowding-out — strong internal validity bought at the cost of external validity, telling you the effect here but nothing about why or where else — because economics is a library of models, not a body of laws, and the scarce skill is diagnosis, picking the model that fits the case. Raj Chetty’s answer is the conceptual trick that lets an imperfect administrative dataset answer a question it was not built for, plus quasi-experiments — grade-level friendship structure, teacher assignment — that manufacture something close to random variation. Brad DeLong works the same uncertainty from the historian’s side: he is only ‘70 per cent sure’ the pre-1870 acceleration would have fizzled into a ‘steampunk economy that looks pretty damn Malthusian’, and candidly ‘runs out of excuses’ for why middle-income traps persist. The shared posture, Studwell aside, is humility: economists are far better at explaining growth after the fact than at causing it.

Where the positions genuinely differ

Is manufacturing special? Studwell and Rodrik build everything on its unique properties — portable technology, unskilled-labour absorption, tradability. Banerjee refuses to privilege it: the crucial feature was a band of middle-skilled work learnable by the poorly educated, and India’s software and business-process-outsourcing boom did the same job with no factories, so there is no reason services cannot substitute — except that automation is now hitting services too.

Does democracy cause growth, or can despotism deliver it? Acemoglu’s evidence says democratisation adds 20–25 per cent to income, and he predicts China’s attempt to pursue frontier innovation while tightening authoritarian control will fall short, because radical innovation needs the individual experimentation a surveillance state suppresses. Yasheng Huang and Dan Wang treat China as the open test: Wang argues East Asian democratisation was always contingent (American occupation in Japan, an assassination in South Korea) rather than the inevitable product of rising incomes, and that the Party has studied the Soviet collapse precisely to avoid repeating it. The forecasts are flatly opposed, and the next two decades adjudicate them.

Can the engine be deliberately built? This is the Studwell–Banerjee axis above, and it determines what a policymaker should even attempt: a forceful industrial strategy, a pragmatic government that ‘asks the private sector what it can do’ and accepts that some bets fail (Rodrik), a relentless attack on confidence and middle-skill employment (Banerjee, Chetty), or a frontier kept dynamic by competition and immigration (Furman).