Abhijit Banerjee on Theory, Practice, and India

Abhijit Banerjee — MIT development economist, co-founder of J-PAL, and freshly minted Nobel laureate — joins Tyler Cowen in Ep. 83, recorded weeks after the prize was announced. The two went to graduate school together at Harvard, and the conversation moves fluidly between the hard limits of development economics, the craft of running randomised experiments, and a long detour through Bengali food, music, and intellectual history. Banerjee is candid about what economists do not know — above all, how to make growth happen.

Key ideas

1. Economists cannot predict or engineer growth — only watch it run out. The one robust finding, Banerjee says, is Robert Solow’s: fast growth slows as a country exhausts its best talent, capital, and locations. What makes growth arrive sooner or later, or stop for decades (Brazil grew fast from 1960 to 1980, then stalled for twenty years), is genuinely not understood. He is sceptical of confident growth recipes drawn from Singapore, Dubai, or Ireland — tiny, idiosyncratic places with very specific political configurations, not replicable templates.
2. The development problem is the petty entrepreneur, and nobody has solved it. Emerging economies are full of subsistence micro-businesses. Universal basic income and microcredit try to make each petty entrepreneur slightly bigger; Banerjee thinks the historical mechanism that actually worked was the opposite — picking a few and letting them grow large enough to employ the rest, pulling people into the labour market. He sees no current model for that transition.
3. A randomised controlled trial (RCT) is a theorist’s instrument, not a statistician’s. An RCT — randomly splitting people into a treated group and an untreated comparison group, so the difference between them estimates the treatment’s effect — is usually sold as a way to get unbiased numbers. Banerjee calls that a red herring. Its real value is that you can keep redesigning the treatment to chase the hypothesis inside your head: why isn’t this working, could it be this instead? The experiment is a way to interrogate a theory in the field.
4. For the poorest of the poor, confidence is the binding constraint. In the multi-country ‘graduation’ programme — a cash grant plus training and coaching — the coaching turned out to be essential (the cash-only version failed in Ghana). The poorest have never succeeded at anything and are treated with contempt; the coaching works by breaking a daunting task into procedural steps (‘here is how you get on a bus’) and signalling that someone believes they can do it. Returns reached several hundred per cent.
5. Liberal modernity in India was born in early-19th-century Bengal, out of contact with the British. A comprador class that worked for the British rapidly absorbed 18th-century European liberal and conservative thought; by the 1820s Bengalis were writing in Bangla about Hume and Rousseau. Kolkata’s Hindu College (1817) was, Banerjee argues, the world’s first purely secular institution of higher education — eight years before Jefferson’s University of Virginia. The Bengali intelligentsia turned sharply left only later, decisively after the 1943 Bengal famine.

Content

Premature deindustrialisation and the vanishing manufacturing premium

Cowen opens with Dani Rodrik’s thesis that emerging economies are prematurely deindustrialising — entering manufacturing just as automation has hollowed out its capacity to build a middle class. Banerjee calls it beyond his competence but says his instincts agree, and adds a complication of his own. Global inequality has shifted spending toward high-quality, heavily designed goods, where the manufacturing labour content is tiny: a worker might earn 40 cents on a $10 toy, the rest going to design, brand, marketing, and safety-checking (often done in Germany, even when the toy is made in China). The earnings that once flowed to manufacturing workers simply are not there anymore.

Pressed on what the positive externality of manufacturing actually is, Banerjee declines to privilege manufacturing at all. The crucial feature, he argues, was a band of middle-skilled work — skills learnable by people with poor education. He sees no reason services could not serve the same function: India’s software and business-process-outsourcing (BPO) boom absorbed relatively low-skilled people who learned English in small-town classes and could carry on a call-centre conversation, no different in kind from upgrading a steelworker to a line supervisor. The catch is that automation is now hitting services too.

Charter cities, entrepôts, and the limits of borrowed institutions

On charter cities — the proposal to let a foreign authority run the law and order of a zone inside a poor country — Banerjee is sceptical not because the idea is bad in principle but because the politics is unstable. A government that hands enforcement to outsiders hands its opponents an easy attack: foreign people are deciding what is legal. And the deeper point: a country that cannot enforce its own laws usually cannot for political reasons, not mere incompetence, and those same reasons make the opportunities for exploitation too tempting.

Cowen counters with history — British Hong Kong flourished, Portuguese Goa did not. Banerjee confesses we understand this badly, and floats geography rather than governance: Hong Kong sat at a superb entrepôt location between Japan, Korea, north China, and India, carrying the opium trade; Goa may simply have been too close to Bombay. There are perhaps only ten to twenty successful entrepôts in all of history — Venice for a while, then not. Whether you can build a durable economy out of an entrepôt is, to him, an open question.

How little economists understand growth

Asked directly how well economists understand growth, Banerjee answers flatly: they do not. The single durable insight is Solow’s — growth tends to run out as the best inputs are used up, and there is nothing tragic about it; China is slowing now, normally. What economists cannot explain is the timing and magnitude: why Brazil grew fast for two decades and then stopped for two more. Cowen presses the apparent counter-evidence — Singapore, Dubai, Ireland all invested in human capital, had the will, ran broadly liberal policies, and grew. Banerjee resists. Singapore in 1960 was a fragile project, lucky in Lee Kuan Yew (it could easily have drawn a leader with crazy views); Dubai works partly as a playground for rich neighbours escaping less liberal societies. These are strange, small, edge-of-other-societies cases, not a transferable recipe.

The graduation programme: cash, coaching, and confidence

The conversation turns to Banerjee’s experimental work — the 2015 Science paper with Esther Duflo, Dean Karlan, and others showing that cash transfers to the very poor, combined with training and coaching, yield returns above 100 per cent, in one case 433 per cent. What is the coaching adding? The cash-only version, run in Ghana, did not work, so the coaching is doing something real. For the poorest of the poor — people who have lived hand to mouth, begged, and been treated with quiet contempt — the binding constraint is confidence. Banerjee relays a participant who had never taken a bus, could not read the route number, and had to be physically shown how to reach the wholesale market.

Cowen asks whether the coaching is essentially a placebo — proof that someone believes in you. Banerjee says it is more than that: it also converts a vague, unlikely proposition (‘go sell things’) into a concrete sequence of steps (‘you go there, you pay this much, they give you something, you bring it back’). Turning a task into a process is itself the intervention. He thinks it is highly scalable — it needs sympathetic, trainable people of the same kind microfinance organisations already employ — and reports that Indian state governments (he was mid-email-thread about Odisha that very day) and some 43 countries are now taking it up.

RCTs as theory in the field, and reforming graduate training

Banerjee gives his clearest statement of method. Randomised controlled trials appeal to him precisely as a theorist. Their point is not unbiased estimation — that, he says, is a red herring achievable many ways — but that you can design and redesign the treatment to follow the hypothesis in your head, keep digging into why something is not working. Theory, in his account, is training for the mind: building ways to fold multiple hypotheses into the sparest possible narrative, learning to ask ‘I think it’s this, but could it be that?’ He credits the Chicago tradition’s interrogative use of theory, whatever its ideological misuses.

This shapes his view of graduate reform. At MIT he sat on a committee that recommended scrapping the general exam in favour of starting research early. The discipline overloads students with ‘this much theory and this much macro is essential’, teaching theorems as descriptions of the world rather than as portable insights. He would teach the insight of why complete markets deliver efficiency not as a theorem about reality but as a lens for spotting what is going wrong — ‘here are 17 examples; find the place where the assumptions are failing’. From his supervisors Eric Maskin and Andreu Mas-Colell he learned to pinpoint the few load-bearing ideas amid the verbiage and ignore the rest.

Bengal: food, music, and the birth of Indian liberalism

A long cultural arc runs through the episode. Kolkata has India’s best sweet shops, Banerjee argues, because a large middle class with surplus met caste constraints that suppressed restaurants — sweet shops became the licensed space where travellers and immigrants of every community could eat freshly made savouries, a quiet mechanism for reconciling Brahminical culture with social mobility. (His aside: rich countries do have fabulous sweets; the United States is the strange exception, and the best Indian food outside the subcontinent is in the UK.)

On music, Bengal produced India’s great instrumentalists — Ravi Shankar, Ali Akbar Khan, Nikhil Banerjee — but never top vocalists; the great classical voices come from the Maharashtra–North Karnataka belt and, differently, the deep South. On intellectual history, he traces the Bengali tradition to the comprador class that absorbed European liberal thought through the British, dating Hindu College (1817) as the world’s first secular university. The crucial turn leftward came late and tragically: the 1943 Bengal famine — three million dead, Churchill diverting Allied grain ships away from India — moved the whole intelligentsia decisively to the left in the 1940s.

Caste, the airline cartel, and lessons from his parents

Cowen raises Acemoglu’s Narrow Corridor claim that the caste system held back Indian growth — yet Pakistan and Bangladesh, without it, are poorer. Banerjee resists a clean answer: we do not know the counterfactual. Caste is awful, but its first-order effects may be less about growth than about serving as an axis of ethnic organisation; without it, India might have fractured along other lines, or unified into a less liberal, anti-minority bloc. The very distrust between scheduled and upper castes may, perversely, supply countervailing power — a stabilising property, at least in the short run.

The conversation closes on lighter and more personal notes — airline loyalty programmes (collusive, he maintains, even as he belongs to two; the industry suffers textbook Bertrand–Edgeworth instability, with Indian carriers like Jet Airways collapsing under too-low prices) — and on his parents. From his mother: to be angry about what is wrong in society, and to interrogate economics that cannot be right. From his father, a liberal in the tradition of Russell and Moore: to unwind a bad argument carefully and respectfully, mindful that the other person has thought about it.

Related

• What Makes Economies Grow — theme; Banerjee is the pole of the engineer-growth axis — economists cannot cause growth, only watch it run out — plus the confidence-as-binding-constraint argument
• Abhijit Banerjee — speaker; MIT development economist, J-PAL co-founder, Nobel laureate
• Tyler Cowen — host
• Daron Acemoglu — MIT colleague; his Narrow Corridor caste argument is put to Banerjee directly
• Daron Acemoglu on the Struggle Between State and Society — companion CWT episode on institutions, colonialism, and the limits of predicting development
• Dani Rodrik on Premature Deindustrialization and Why the World is Second Best, at Best — the premature-deindustrialisation thesis Cowen opens with
• Brad DeLong on Intellectual and Technical Progress — companion CWT episode on long-run growth and economic history
• Joe Studwell on Africa, Asia, and What Development Actually Requires — related episode on what development actually requires