Lant Pritchett on Poverty, Growth, and Experiments
What this covers
Lant Pritchett and host Russ Roberts examine why poverty persists globally and which interventions matter most. Pritchett argues that poverty is fundamentally a problem of living in low-productivity economies rather than of individual deficiency, and this distinction reshapes which solutions deserve attention. The conversation moves from the scale of historical poverty reduction—comparing what happened in China, India, Indonesia, and Vietnam—to the specific mechanisms development economists use to study interventions, with Roberts probing tensions between rigorous causal inference and the messier question of what actually lifts nations out of destitution.
Pritchett develops several overlapping critiques. He documents the Place Premium: observationally equivalent Nigerians earn roughly 16 times more doing the same work in the US, a gap that persists even after correcting for selection. He challenges what he calls the Perils of Partial Attribution—the field's bias toward small, measurable programmatic effects like deworming over large systemic factors like market-oriented reform, which cannot be easily isolated in an experiment but explain nearly all historical poverty reduction. Migration emerges as the single largest poverty intervention, yet it remains politically unthinkable; Michael Kremer's influential deworming study is debated, with larger-scale replications showing smaller effects. Against this, Pritchett argues for investing in the intellectual Mound of Gravel—the accumulated consensus of research and ideas that shifts policy beliefs—citing China's post-1978 market turn and India's 1993 reform acceleration as examples where collective intellectual work on economic direction yielded trillions in GDP gains that no single study can claim credit for. He also questions the dollar-a-day poverty metric itself as an arbitrary low bar that obscures the full scope of deprivation.
Pritchett argues that world poverty is overwhelmingly a problem of low-productivity places, not deficient individuals, so the highest-return interventions are systemic (economic growth, market-oriented reform, migration, and building the intellectual 'mound of gravel' that shifts policy) rather than easily-attributable programmatic fixes like giving people chickens or running cash-vs-chicken experiments.
- People are poor because they live in low-productivity economies; the same person earns ~16x more doing equivalent work in the US vs Nigeria.
- Nearly all historical poverty reduction (China, India, Indonesia, Vietnam) came from shifts to market-oriented growth, not poverty programs.
- Obsession with directly-attributable causal effects ('Perils of Partial Attribution') biases research toward small, measurable questions and away from the large systemic ones that actually matter.
Market-oriented reforms and economic growth are the primary drivers of poverty reduction worldwide.
- China's post-1978 shift to a more market-oriented economy under Deng Xiaoping added about $11 trillion in GDP relative to the counterfactual; even if the global development industry—by creating the ideas and evidence influencing that decision—contributed only one tenth of one percent of that gain, it was worth it by an order of magnitude.
“China has added $11 trillion dollars in additional GDP, relative to the counterfactual of what we would have predicted for China had China continued.”
- Essentially none of the poverty reduction in Indonesia from 1965-1996, and only a very small fraction in Vietnam where poverty fell for over half the population to under 10%, was due to poverty programs—so imagining that the highest returns to poverty reduction come from more effective programs, when none of the actual variation is explained by them, is a fundamentally misguided 'drunk-under-the-headlight' approach to science.
“We know, for sure, 100% that none of the poverty reduction in Indonesia from 1965-1996 was poverty programs.”
- Nearly all reduction in dollar-a-day poverty occurred because people moved from low-growth to high-growth economies via market-oriented reforms that raised overall productivity, the poor's productivity, and their wages—so economic growth is the key to reducing both low-bar and high-bar poverty, and is the only way to reduce high-bar poverty.
“economic growth is in fact the key to reducing even low-bar poverty. And unambiguously, unquestionably the only way to reduce high-bar poverty.”
- India's two growth accelerations, dated econometrically to 1993 and 2002 following its shift from an inward, regulated economy toward a market-oriented one, together added at least $2 trillion of GDP that would not have been produced under business-as-usual growth—making investment in the conditions that enabled that reform vastly higher-return than giving Indians chickens in 1991.
“Those two reforms combined added at least $2 trillion of GDP... that wouldn't have been produced, counterfactually, had they maintained business-as-usual growth”
Research and ideas collectively shape policymaker beliefs more powerfully than direct program attribution.
- The free trade consensus was a deliberate, successful 20th-century social movement: thirty years ago countries like India banned consumer imports outright, and advocates consciously used every lever to change beliefs and swept the board—a transformation that no single study can be credited with but which collectively had a huge impact, and which recent revisions only roll back about 2% of the way.
“There were a group of us that were committed to free trade... And we consciously used every lever at our disposal to change people's ideas about free trade. And we won. We swept the board.”
- Individual economic studies each look trivial in isolation, but the professional consensus and intellectual milieu they collectively build—the 'mound of gravel'—becomes genuinely influential on policymakers; so investing in research and ideas that shape what the world believes is far higher-return than directly-attributable programs, even though no single study's effect can be isolated.
“you do all of this kind of work not because you expect the individual report that's written to make a difference, but the overall milieu and the professional consensus that arises about general directions creates ideas about the world that themselves become influential.”
- The obsession with directly-attributable causal effects—what Pritchett calls the Perils of Partial Attribution—systematically biases research toward small, easily-measured questions and away from the large systemic factors that actually explain poverty reduction, like focusing on a basketball player's points scored while ignoring his defense, which can make a high scorer a net-negative contributor to team wins.
“moving too much to the attribution model--what am I doing for poverty that I can attribute directly to the actions of this individual?--is like focusing exclusively on points scored of an individual player and not thinking about, 'What are you contributing to the team wins?'”
- The central question should be how to get a country from a 320 to a 500 in learning outcomes, and answering it requires having defensible presumptions about direction—like Smith's presumption of liberty or 'teaching at the right level'—that produce cumulative huge effects, rather than treating directly-attributable measurability as the criterion for what's important; success is the issue, not being the high scorer.
“I think that's the huge mistake: is we're running into causal, attributability as the issue versus success as the issue. Success is the issue. We should be focused on success, not on being the high scorer.”
Systemic factors like location and economy determine poverty; individual pathology and small interventions do not.
- Most high levels of poverty reflect systemic rather than personal pathology: there aren't poor people so much as people in poor places. Nigerians measured as poor have no special poverty-creating characteristic; they are poor because they apply their labor, skills, and capital in a low-productivity environment determined by the economy's general productivity factor.
“Mostly in the world there aren't poor people. There are people in poor places.”
- The 180-point gap between Indian (~320) and US (~500) student achievement on internationally normed assessments cannot be meaningfully closed by deworming: the causal chain (fraction of kids with worms × reduction from deworming × incremental score gain) yields an effect closer to 1 point than 10 or 100, so research should focus on first-order systemic features—how teachers are hired, assessed, and perform, and curriculum fit—rather than easily-attributable interventions.
“the net impact is going to be fraction of kids actually suffering from worms, times reduction in fraction of kids suffering from worms from deworming, times the incremental score of an Indian kid with no worms over an Indian kid with worms... a number that's going to be more like 1 than it's going to be like 10”
- The dollar-a-day poverty standard, invented in 1991 for a World Bank report, sets the bar so low that while everyone below it is unambiguously poor, it wrongly excludes the many people just above it who are still legitimately poor, biasing global attention toward an arbitrary low-bar definition with no fundamental legitimacy in economics or politics.
“if you are at $1.01 a day, or $1.02 a day, or even $2.00 a day, is there any meaningful sense in which you can be said to be not poor?”
- On every measure of human wellbeing—education rising from 2 to 7 years in developing countries, plummeting infant mortality, sustained ~2% growth making the world enormously richer, and absolute poverty plummeting—the 60 years from 1950 to 2010 were the best in human history by a multiple, coinciding with the self-conscious global project of promoting development.
“The years from 1950-2010 have, on every single measure of human wellbeing, been the best 60 years in the history of man by a factor of multiple.”
- The money given via foreign aid was not the causal mechanism that produced development success—evidenced by China and India receiving little aid yet succeeding while heavy aid recipients did far worse—rather it was the broader 'team development' project of ideas, research, and an open trading order that mattered.
“I don't think the money given via aid was particularly the causal mechanism that caused success.”
- On the 2009 PISA exam, the two Indian states tested (chosen as the country's best) scored around 320 against an OECD mean of 500 with a 100-point student standard deviation, meaning the average Indian 8th-grader would land in the bottom 1-percentile if placed in a Danish or US school—and this almost certainly overstates India-wide performance.
“the average student in India in 8th grade would be in the bottom 1-percentile if they were suddenly put in a Danish or U.S. school. Unbelievably low”
Migration from low-productivity to high-productivity places is the largest untapped lever against poverty.
- Letting people move from low-productivity to high-productivity countries is unambiguously the largest and most reliable way to attack global poverty—illustrated by the fact that 82% of all non-poor Haitians live in the United States—but it is largely ignored for political reasons.
“82% of all Haitians who aren't poor are living in the United States. So, the route out of poverty for a nation is to get out of Haiti into a high-productivity environment”
- Comparing observationally equivalent Nigerians—same birthplace, same education—working in the US versus Nigeria, a low-skilled worker's wage rises roughly 16-fold simply by being in a high-productivity environment, and after six different methods of correcting for selection, at most about 25% of the gain is selection, leaving a massive residual attributable to place.
“that wage, for a low-skilled person, somebody with less than high-school education, it's 16 times. So, your wage goes up 16-fold.”