Abhijit Banerjee
What this covers
Abhijit Banerjee sits with Tyler Cowen for a wide-ranging conversation spanning economic development, history, Indian intellectual life, and the craft of economic thinking itself. The discussion pivots on a core claim: that economists have severely misunderstood the engines of growth and the pathways out of poverty. Banerjee argues that while economists grasp that growth eventually slows, they cannot reliably explain what starts it, sustains it, or halts it abruptly—Solow's observation about diminishing returns is the only consistent insight. The conversation moves from this theoretical puzzlement into the practical findings from field experiments: RCTs reveal their power not as a method for unbiased estimation but as a tool for iteratively testing hypotheses against context, and coaching interventions for the poorest show that confidence and procedural knowledge matter as much as capital itself.
The substance splits into distinct registers. Banerjee challenges several orthodox positions: that manufacturing holds a special developmental role (arguing services could serve the same function), that microcredit and unconditional cash transfers alone lift people out of poverty (they do not without accompanying coaching), and that small success stories like Singapore prove scalable models (they are fragile and non-replicable). He moves into Indian history, tracing how Kolkata became an intellectual center through the work of figures like Ram Mohan Roy, and how caste systems may function as organizing structures with countervailing effects. He discusses why charter cities fail, how airlines face structural instability through Bertrand-Edgeworth competition, and why Indian vegetables taste better. Throughout, he articulates a vision of theory as portable insight—a toolkit for spotting where assumptions break rather than a collection of theorems—and defends a reformed approach to training economists that prizes early research and applied reasoning over mandatory comprehensive theory.
Banerjee argues that economic development is poorly understood by economists—growth is unpredictable, manufacturing's special role is overstated, and what matters is accessible mid-skill pathways—while randomized trials and confidence-building interventions reveal practical, context-specific mechanisms that broad theories miss.
- Economists understand that growth slows but not what makes it start or stop, so growth prediction is unreliable.
- RCTs are valuable not for unbiased estimates but for iteratively testing hypotheses against context.
- Coaching the poorest builds confidence and procedural knowledge, not just capital, which is why cash alone fails.
Economic growth itself remains largely unexplained despite Solow's insight that it exhausts natural limits.
- Economists do not really understand economic growth; the only consistently borne-out insight is Solow's—that growth tends to run out once the best talent, capital, and locations are used up—but what makes growth slow sooner or later, or stop unexpectedly (as Brazil's did after 1980), is not understood, and growth prediction is unreliable.
“I think we’re lousy at predicting growth in general.”
- Apparent growth successes like Singapore and Dubai do not prove we understand growth; they are tiny states with very specific, fragile political configurations (Lee Kuan Yew's commitment, or proximity to illiberal societies needing a 'playground'), making them poor and non-replicable models at scale.
“Singapore was a very, very fragile project in 1960 and one that, I think, they were lucky to have him as the dictator.”
- The 1999 QJE paper with Aghion and Piketty was an early precursor of secular-stagnation theories that never gained traction with macroeconomists because it used a spare theorist's model lacking the 'bells and whistles' that mainstream macro modelers expected.
“It never flew because I think people really wanted the bells and whistles that we didn’t really provide.”
- Pakistan's economy is doing fine, growing at around 6 percent for many years, and with Indian growth faltering it may converge—6 percent growth is nothing to dismiss, though the causes of Pakistan's growth are not well understood.
“After all, it still grows at 6 percent, which is not nothing... For many, many years, it’s been growing at 6 percent.”
Economics should teach portable thinking tools to interrogate assumptions rather than memorize theorems or mandated theory.
- The real value of randomized controlled trials is not getting unbiased estimates—a 'red herring' achievable other ways—but the ability to iteratively design and vary the treatment to test exactly the hypothesis in the researcher's head, making RCTs appealing to a theorist as a tool for interrogating why something does or doesn't work.
“it’s really not about getting unbiased estimates. That’s sort of a red herring.”
- Graduate economics training should be reformed by dropping general exams and encouraging early research, because overloading students with mandatory theory and macro 'stuns' them so they don't learn; the goal should be portable insights students can immediately apply, taught by having them spot failing assumptions across examples rather than memorize the edifice.
“I was on a committee which recommended that we get rid of the general exam and we just, instead, have much more encouragement to start doing research early.”
- Economic theory is best understood as a toolkit and training for the mind—generating portable insights to interrogate multiple hypotheses—rather than place- and time-independent theorems about the world; teaching should make students spot where assumptions fail across examples (e.g., understanding why complete markets deliver efficiency to see what's going wrong) rather than memorize theorems.
“what’s the best thing with economics is that it gives you a toolkit. You interrogate things”
- The core skill learned from theorist supervisors (Maskin and Mas-Colell) is the discipline to pinpoint the few key ideas amid verbiage—finding the one place an argument fails or what gives it leverage—rather than getting lost in often man-made, unnecessary complexity.
“Just have the discipline to pinpoint. Don’t get lost in the complexity. The complexity is often man-made and sometimes unnecessary.”
- The airline industry's first-order problem is Bertrand-Edgeworth competition: firms pay large fixed costs then compete with very low marginal costs, which is known to be unstable—players drive each other out and prices then rise—visible now in India where airlines like Jet Airways and Kingfisher have gone out of business, leaving roughly three and a half airlines.
“The problem with the airline industry is also what, in economics, we were taught as Bertrand-Edgeworth competition. You pay your fixed cost and then you compete. It’s known to be unstable.”
Coaching poor entrepreneurs builds confidence by breaking impossible tasks into concrete procedural steps yielding high returns.
- Coaching the poorest is more than a placebo of conveying confidence; it works by breaking unfamiliar, seemingly impossible tasks (like buying wholesale and reselling) into concrete procedural steps—how to get on the bus, pay, collect goods, bring them back—making an unlikely proposition feel achievable.
“Turning things into a set of processes is important. Otherwise, it looks like an unlikely proposition that I can do it.”
- Cash transfers to the poorest of the poor combined with training and coaching yield very high returns (over 100% up to 433% in a six-country study), but the same transfers without coaching did not work in Ghana, indicating the coaching itself is doing something essential—largely building confidence in people who have never succeeded at anything.
“we’ve actually done it without the coaching — in Ghana, and it doesn’t work. So we are reasonably confident that the coaching is doing something interesting.”
- The coaching/cash-transfer 'graduation' approach is scalable because it requires only sympathetic NGO-level staff at the same skill level as microfinance workers, and it is being adopted across many countries and Indian state governments (e.g., Orissa), with reports of experimentation in dozens of countries.
“Many state governments in India now want to just do it... we were a part of a long email exchange on exactly how it should be done in the state of Orissa”
Manufacturing cannot generate large worker earnings because global income concentrates in design, branding, and distribution.
- The development benefit usually attributed to manufacturing actually comes from offering a set of middle-skilled jobs whose skills are learnable by people without good educations; the same role could potentially be played by services (e.g., Indian BPO and software work), not specifically manufacturing.
“it was there’s some skills involved, and the skills can be learned, and the skills were available to people who didn’t start with having an excellent education.”
- Because so much global income is concentrated and spent on high-quality, heavily designed and brand-marketed goods, the manufacturing labor and capital cost is a tiny fraction of final price (a worker may get 40 cents of a $10 toy), so manufacturing can no longer generate large earnings for workers in emerging economies.
“if you’re a worker, you’re getting 40 cents for a $10 toy or something, and that’s just not going to make a lot of rich people or even middle-class people.”
- Indian vegetables taste better largely because of short supply chains (produce picked the same morning and sold locally) and the fact that food is not bred for durability, whereas US tomatoes are designed for durability, which sacrifices taste.
“In the US, tomatoes are designed for durability, and that takes away from other things.”
Charter cities and entrepots depend on scarce locations and fragile politics, making them non-replicable policy models.
- We do not understand why some charter cities/entrepots like Hong Kong succeeded while others like Goa failed; success may simply reflect a scarce great location (the world can only support a limited number of entrepots) rather than any replicable policy, leaving open whether economies can actually be built out of entrepots.
“There are only 10 examples of very successful entrepots, or maybe 15 or 20.”
- Charter cities are fragile because handing over law enforcement to a foreign enforcer is an easily exploited political target; the very reason a country cannot enforce its own laws is usually political rather than mere incompetence, so outsiders enforcing them is internally contradictory and unstable.
“This is politically extremely unstable, easy to attack, very visible targets.”
Social structures like caste and immigration patterns shape economic institutions without directly determining growth trajectories.
- Kolkata's sweet shops arose because caste rules limited restaurants and made eating out difficult, while a permitted line allowed certain castes to make sweets and savories; these shops became the core mechanism for reconciling Brahminical food cultures with the social mobility of immigrants flowing into a large city.
“it was actually the core mechanism for reconciling Brahminical cultures of different kinds with a certain amount of social mobility.”
- The Bengali intellectual tradition arose because Bengalis (and Parsis) were the comprador classes working with the British, who by around 1800 were rapidly absorbing English and Scottish Enlightenment writers; the resulting liberal tradition (epitomized by Ram Mohan Roy, a Benthamite contemporary) moved leftward partly in reaction to reactionary conservative pushback.
“what happened then was very much people absorbing English 18th-century liberal and conservative discourses, using them.”
- The 1943 Bengal famine, in which roughly three million died amid a callous government response and Churchill diverting American grain ships away from India, was the defining event that sharply moved the previously liberal Bengali intelligentsia to the left starting in the 1940s.
“the Bengal famine just completely moved everybody to the left because the Bengal famine in Bengal is seen as being egregious completely, a callous government”
- The economics of Indian classical music are sustainable through concertizing—top performers' fees have risen and they are in demand—supplemented by Bollywood, which is a generous and reliable client for good classical musicians; the family-based father-to-son training tradition is now an anachronism replaced by better training academies.
“if you look at the top echelon of these people, their fees have gone up a lot, and they’re in demand.”
- There is amazingly good Bengali novel-form literature and poetry from the 1920s to 1970s (Bibhutibhushan Bandyopadhyay, Tarasankar Bandyopadhyay, Manik Bandopadhyay, Satinath Bhaduri) that is translatable but neglected because there has been insufficient incentive to attract talented translators.
“Given incentives, there is just so much great literature.”
- What is now Presidency University, established in Kolkata in 1817 as Hindoo College, was the first purely secular institution of higher education in the world—funded by the Hindu community but open to all faiths with a Christian (Derozio) as its first famous professor—predating Jefferson's University of Virginia (1825) by eight years.
“The one fact that I like to emphasize is the first, I think, purely secular institution of higher education in the world is in Kolkata in 1817.”
- While the caste system is awful, its first-order effects may not be directly on economic growth; it may instead function as a form of ethnic organization, and counterfactually, divisions like Scheduled Castes distrusting upper castes could provide countervailing power that keeps India more liberal than a unified-Hindu polity organized along anti-minority lines.
“I don’t know whether I would think its first-order effects are on growth directly or because they just happened to be a way of ethnic organization.”
- A key intellectual virtue, learned from his father, is to listen carefully to others' arguments, assume they have thought them through, and respectfully unwind them—being ruthless about bad arguments but in a funny, nonconfrontational way.
“before you dismiss them, be mindful that they have thought about it. Be respectful of what their argument is and kind of unwind it carefully.”
- Belonging to consumer loyalty programs is rational in equilibrium—if you belong to it in equilibrium, you should belong to it—and the welfare losses from airline loyalty schemes are not necessarily large.
“the fact that I belong to it in equilibrium — I should belong to it.”
- The US is exceptional in how rarely one finds excellent sweets and bakery goods, whereas France, Italy, and Spain consistently produce wonderful pastries even in no-name village bakeries at trivial price premiums.
“the sweet shop there produces better sweets of different kinds than any place in Boston that I would go.”
Petty entrepreneurs historically advanced through rare individual success and employment coordination, not microcredit or UBI expansion.
- The microcredit/UBI literature is stuck trying to make petty entrepreneurs slightly bigger, but historically petty entrepreneurs were absorbed into the economy by a few becoming very successful and employing many others (often via brand coordination, as with Benetton in Italy); there is no current mechanism for moving the smallest entrepreneurs out into the labor market.
“the way petty entrepreneurs have been absorbed into the economy historically has been by making some people being very successful, employing lots of people and creating large enterprises.”