For seventy years one subfield of economics has been trying to answer a single question: why are some countries poor, and how do they stop being poor? It has answered that question twice over, in opposite registers, and the swing between the two answers is the story of this chapter. The first register was grand structural theory — the conviction that a poor economy has a distinct shape you can model and engineer, that development is a transformation a planner can see coming and help along. The second was piecemeal credible measurement — the conviction that the only honest knowledge about development comes from testing specific interventions one at a time, and that the grand theories were ambition without evidence. The field swung from the first pole to the second across the second half of the twentieth century, and the swing left a discipline that knows far more than it used to about which intervention works and far less than it once claimed about what makes a whole economy grow.
The arc runs through five movements. It opens with the structuralist founding — W. Arthur Lewis's dual-sector model and Rostow's stages of growth, the moment economics decided poor economies needed their own theory. It moves to the big-push tradition, where development became a coordination problem markets could not solve alone. It turns to the periphery's answer — Prebisch, dependency theory, and import-substitution industrialization — and then to the neoclassical counter-revolution that judged that program a failure and prescribed openness, only to over-prescribe in turn. It pauses on Sen, who relocated the question itself from income to freedom. And it ends with the randomized-trial turn that reinvented the field's method and, by design, declined the question Lewis had started with. Each pole over-promised; each correction carried real gains and real costs; and the dispute the corrections left behind — whether the new methods can answer the old question at all — is genuinely unsettled, which is exactly why the structural tradition has not died.
A poor country is not a rich country with less money in it. That sentence is obvious now; it was a founding insight when W. Arthur Lewis built a model around it. Before the 1950s, the few economists who thought about poor countries at all tended to apply the standard apparatus of rich-economy theory with the magnitudes turned down — less capital, lower wages, the same equilibrating markets. Lewis's claim, and the claim of the subfield that formed around him, was that this missed the structural fact about a poor economy: it is not one labor market but two, joined awkwardly, and growth is the story of one absorbing the other.
Lewis's 1954 paper, "Economic Development with Unlimited Supplies of Labour," set out the dual-sector model that became the field's seminal apparatus. A developing economy, on this account, has two sectors. A traditional or subsistence sector — peasant agriculture, household production, informal urban services — holds the bulk of the population and is crowded with labor. It is so crowded that some of that labor is surplus: workers whose removal would leave the sector's output essentially unchanged, because their marginal product is at or near zero. A second worker on a small family plot, a third hawker on the same corner, adds almost nothing to what the sector produces. Alongside this sits a modern capitalist sector — factories, plantations, mines — small at first but operating on different principles: it hires labor for a wage, earns profits, and accumulates.
Growth, in this model, is the modern sector drawing labor across the divide. Because the traditional sector holds surplus labor, the modern sector can hire as many workers as it wants at a roughly constant wage — a wage set a little above subsistence, enough to pull a worker off the farm but no higher, because there is always another worker behind the first. The modern sector's profits are reinvested, its capital stock grows, it hires more labor at the same flat wage, and the cycle repeats. The crucial feature is that the wage does not rise while the surplus lasts. Capitalists capture the gap between what the new workers produce and the constant wage they are paid, plow it back into more capital, and absorb still more labor. The traditional sector shrinks; the modern sector grows; and the process runs until the surplus is exhausted. That exhaustion is the Lewis turning point — the moment the modern sector has soaked up the traditional sector's slack, after which any further hiring competes for labor that is now genuinely scarce, wages start to rise, and the two sectors merge into a single integrated labor market. Development, on this picture, is precisely the passage to that point.
The reason this counted as a founding move, rather than one more growth model, is the contrast with the apparatus economics already had. The aggregate growth model that chapter 9 walks — Solow's — treats the economy as a single sector with one production function combining capital and labor, and asks how saving, investment, and technical change move output per worker over time. It is a powerful model, and the development field defined itself partly against it. Solow's single aggregate sector has no place to put the dualism: it cannot represent surplus labor whose marginal product is zero, because in a smoothly differentiable production function every worker has a positive marginal product. What Lewis saw was exactly what Solow's frame abstracted away — that a poor economy's defining feature is not a low level of the same variables but a different structure, two sectors obeying different rules, with growth as the migration between them. The formal statement of structural transformation and the formal Solow benchmark both live in Economics chapter 20 and chapter 13; here the point is the intellectual one, that taking the dualism seriously is what made development economics a subject.
The model's assumptions are also where its critics found their purchase, and the criticism is worth marking because it tracks the chapter's larger argument about ambition and evidence. The constant-wage premise — that the modern sector can hire indefinitely at a fixed subsistence-plus wage — held only as long as the surplus-labor reservoir was real and deep, and empirical work questioned both halves: whether marginal product in poor-country agriculture was actually zero (Theodore Schultz argued it was not, that peasant farmers were "poor but efficient" and allocated their labor sensibly), and whether modern-sector wages stayed flat (in much of the developing world they rose well before any plausible turning point, pushed up by minimum-wage laws, urban labor regulation, and union pressure rather than by the exhaustion of rural surplus). The model thus described a clean mechanism that the messy data only partly obeyed — a pattern that recurs through this chapter, in which the structuralist tradition's most elegant constructions are the ones that fit the world least exactly. What survived the criticism was not the specific predictions but the structural frame: the insistence that a developing economy is a transition between qualitatively different sectors, and that the transition, not the accumulation of a single homogeneous capital stock, is what development is.
Drive the labor transfer. Drag workers out of the traditional sector and into the modern one and watch the modern-sector wage stay flat — at subsistence-plus — for as long as surplus labor lasts, then kick upward once the surplus is exhausted. The kink you produce is the Lewis turning point; the flat stretch before it is the whole structural claim. Then flip to the contrast view to see why a single aggregate sector — Solow's frame — cannot show any of this.
Figure 16.1 (interactive). The modern-sector wage as labor transfers from the traditional sector. It stays flat at subsistence-plus while surplus labor lasts, then rises past the Lewis turning point. Drag the transfer slider; the turning point shifts with the capital-stock slider. The contrast view replaces the two panels with a single aggregate sector that cannot represent the dualism.
As long as the farm has labor it can spare for free, the factory never has to bid the wage up — there is always another worker behind the first. So growth is the factory eating the farm at a constant wage, and the capitalist pockets the gap and reinvests it. Only when the farm runs out of spare hands does the factory have to compete for labor, and the wage finally climbs. A single-sector model has no farm to drain, so it cannot tell this story at all.
The closed-form dual-sector dynamic — the constant-wage labor-supply curve, the turning point, and the contrast with the aggregate production function — is developed in Economics chapter 20.2, with the Solow benchmark in chapter 13. This widget computes a stylized wage path to make the turning point drivable, not Lewis's full derivation.
Around Lewis the era assembled its confident teleology. W. W. Rostow's The Stages of Economic Growth (1960) laid out a sequence every society was held to pass through — traditional society, the preconditions for take-off, the take-off itself, the drive to maturity, and the age of high mass consumption — with the "take-off" as the decisive moment when growth becomes self-sustaining. Rostow's stages have not aged well as a theory; the sequence is too tidy and the historical record too lumpy. What endures about the book is its subtitle: A Non-Communist Manifesto. Development economics was born inside the Cold War, and the question of how poor countries grow was inseparable from the question of which system — the Western market path or the Soviet planned path — would deliver that growth first. Rostow wrote a sequence that ended in mass consumption rather than in revolution, and he meant the contrast. The ideological stakes shaped which models got built and which got funded, and they run under the structuralist-versus-neoclassical quarrels that follow.
Lewis received the Nobel in 1979, the first — and for a long time the only — development economist so honored, and the first Black laureate in any of the sciences. The prize marked the field's arrival as a recognized branch of the discipline rather than a peripheral concern. Where Lewis sits relative to the rest of the lineage is on the development-economics cluster of the intellectual-history timeline. The structural conviction he founded the field on — that a poor economy has a shape, and that development is a transformation of that shape — sets up the next move, which asked what happens when the market cannot accomplish the transformation on its own.
You just drove the model the whole lineage starts from. The rich/poor question doesn't get answered here — it gets given its first serious shape.
The lineage starts here. Lewis saw a poor economy as structurally dual — not a small rich economy but two sectors, growth being the modern one absorbing the traditional. That structural answer is the first of several this walkthrough holds against each other; what makes a whole economy develop is the question the field will spend seventy years circling.
Development economics is the growth lineage's branch into the poor-country question.
This is where the growth question turns toward poor economies specifically. From Lewis's structural transformation runs a thread to the question of what makes a whole economy develop — the question the RCT turn at the end of this chapter most conspicuously declines to answer, which is why the structural and endogenous-growth branches stay live.
Rostow's subtitle is the giveaway.
Rostow's Stages was subtitled "A Non-Communist Manifesto" — development economics was a Cold-War battleground over which system, market or plan, delivered growth first. The structuralist-versus-neoclassical quarrels later in this chapter carry that ideological charge under the technical argument.
Build one factory in a country with no customers and it fails. Build ten factories that are each other's customers, and they all succeed. So the market, building one factory at a time, builds none of them. That is the intuition behind the period's most distinctive contribution: the idea that development is, at bottom, a coordination problem — a problem about getting many investments to happen together when none of them is profitable alone.
Paul Rosenstein-Rodan gave the argument its canonical form in 1943, in a paper on the industrialization of Eastern and South-Eastern Europe, and named the solution the big push. His illustration was a shoe factory. A single entrepreneur who builds one shoe factory in a poor agrarian region gains nothing: the new workers spend their wages on food and many things, but only a fraction on the factory's own shoes, so it cannot find enough demand to be profitable. Now suppose a hundred factories open at once, across complementary industries. Each factory's workers become customers for the others' products. The demand that no single factory could generate for itself is generated jointly, by the whole bundle of investments, and the bundle is profitable even though no piece of it is profitable alone. The implication is that a market left to itself, deciding one investment at a time on each investment's own merits, will build nothing — and that a coordinated, simultaneous push across complementary sectors is what escapes the low-development equilibrium.
The argument had a postwar urgency that is easy to miss now. Rosenstein-Rodan wrote for a moment when the reconstruction of war-shattered economies and the development of newly independent ones looked like the same engineering problem, and when the apparent success of Soviet planning gave the coordinated-push idea a rival model with real momentum. If a market could leave a poor economy stuck in a low equilibrium, then a state that could see the whole board and move many pieces at once held a genuine advantage — and the big-push logic supplied the theoretical license for the large, state-directed investment programs that development agencies, national planning boards, and the early World Bank built through the 1950s and 1960s. The idea was not a curiosity; it was, for two decades, the intellectual backbone of how the rich world proposed to develop the poor one.
Ragnar Nurkse gave the same insight a sharper edge with the vicious circle of poverty and the case for balanced growth. A poor country is poor, Nurkse argued, in a self-reinforcing loop: low income means low saving, low saving means low investment, low investment means low productivity, and low productivity means low income again. The circle holds the economy in place. The way out is to invest across many sectors simultaneously, in balance, so that each industry's output finds a market in the others' rising demand — a coordinated assault on the circle at many points at once, because attacking it at any single point leaves the others to drag the effort back down. The poverty trap and the big push are two faces of one idea: that a poor economy can sit in a low equilibrium it cannot escape by incremental, piecemeal change, and that escape requires a large, coordinated leap.
Albert Hirschman dissented, and the dissent became the era's signature internal quarrel. His Strategy of Economic Development (1958) argued that the balanced-growth prescription asked the impossible: a country too poor and too administratively thin to coordinate one industry could hardly coordinate a hundred at once. If a government could orchestrate a simultaneous balanced push across all sectors, Hirschman observed dryly, it would not be underdeveloped in the first place. His counter was unbalanced growth: deliberately create disequilibria. Invest in a few leading sectors, let the shortages and bottlenecks they generate pull further investment along behind them. The mechanism was linkages. A backward linkage is the demand a new industry creates for its inputs — a steel mill that needs ore and coke and machinery pulls investment into those upstream supplies. A forward linkage is the supply a new industry offers to downstream users — cheap steel invites the fabrication and engineering industries that consume it. Hirschman's claim was that a poor economy should pick the sectors with the strongest linkages and push there, letting the economy's own induced-investment pressures do the coordinating that no planner could do deliberately. Balanced growth trusted the plan; unbalanced growth trusted the disequilibria. The quarrel between them was the structuralist tradition arguing with itself about how much coordination a poor state could actually perform.
Set how many complementary sectors invest at the same time. Below a coordination threshold each sector loses money — no customers yet — and the economy sits in the low (trap) equilibrium. Push past the threshold and each sector becomes the others' customer, profits go positive, and the economy jumps to the high equilibrium. The two stable resting points with an unstable threshold between them are what make the poverty trap a trap: you cannot inch out of it, you have to leap.
Figure 16.2 (interactive). Per-sector profit against the number of sectors investing together. Profit is negative below the coordination threshold and positive above it; the economy's equilibrium falls to the low (trap) state below and locks into the high state above. Drag the count of coordinated investors across the threshold; the complementarity slider moves where the threshold sits.
A factory needs other factories' workers as its customers. With too few building at once, nobody has customers, everybody loses money, and the economy falls back to where it started — the low equilibrium. With enough building together, each is the others' market, profits appear, and the economy holds at the high equilibrium. The threshold between them is the gap a coordinated big push has to clear in one jump; a market deciding one factory at a time never clears it.
The multiple-equilibria formalization of poverty traps and the big push — complementarities, the coordination failure, and the basin structure — is in Economics chapter 20.3. This widget computes a stylized threshold payoff to make the two equilibria drivable, not the full model.
The poverty-trap idea has a long afterlife. The big-push tradition made the strongest pre-empirical case that markets can leave a poor economy genuinely stuck — not merely growing slowly but locked in a low equilibrium from which no incremental investment escapes. That is a powerful claim, and it justified ambitious, coordinated, planned intervention. It is also exactly the claim the randomized-trial generation of the 2000s most directly contests, asking whether the dramatic poverty traps the theory posits actually exist in the data, or whether poor people and poor economies respond to incremental change after all. That argument arrives later, and the coordination logic set out here is the structural pole it contests — which is why the case for it has to be stated at its strongest first.
The modern debate has an ancestry, and it runs through this section.
The intellectual ancestry of today's industrial-policy revival is the big push and, just ahead, the developmental state — the case that some coordination markets cannot do needs a visible hand. The big push is the lineage anchor: if complementary investments are jointly profitable but separately not, a coordinated push is the escape, and the modern argument is whether states can still pull it off.
Ricardo's theory of comparative advantage, which chapter 3 walks, promised that every country gains by specializing where it is relatively most efficient and trading for the rest. Raúl Prebisch looked at Latin America specializing in coffee and copper and tin, exactly as the theory advised, and saw the same logic running a different way: a machine for transferring the periphery's productivity gains to the industrial center. The structuralist challenge that came out of the global South was not a rejection of the trade logic but an inversion of it — the claim that, for a primary-commodity exporter, comparative advantage could be a trap rather than a gift.
The mechanism is the Prebisch-Singer thesis, stated independently by Prebisch and Hans Singer around 1949–50: the terms of trade — the ratio of the price of a country's exports to the price of its imports — tend, for primary-commodity exporters, to decline against manufactured goods over the long run. If the price of the coffee and copper a country sells falls relative to the price of the machinery and manufactures it buys, then the country must export ever more to import the same, and its productivity gains in the export sector flow abroad as cheaper commodities rather than staying home as rising incomes. Prebisch posited two reinforcing reasons. On the demand side, as world incomes rise, spending shifts proportionally more toward manufactures and services than toward food and raw materials — an income-elasticity asymmetry that depresses primary-commodity prices over time. On the supply side, the labor-market structure differed between center and periphery: in the industrial center, strong unions and tight labor markets let productivity gains be captured as higher wages and held in prices, while in the periphery, surplus labor (the Lewis reservoir of section 16.1) meant productivity gains were competed away into lower prices. The result was a structural drift of the terms of trade against the periphery, reproducing its disadvantage through the very act of trading.
It is worth being honest about the empirical status of the thesis, because the chapter's argument does not depend on settling it. The long-run terms-of-trade record for primary commodities is genuinely mixed — some series show the predicted decline, others show no clear trend, and the answer depends heavily on which commodities, which period, and how quality change in manufactures is handled. What matters for the lineage is not that the empirical claim is settled (it is not) but that the mechanism Prebisch posited was a serious one, with a real grievance behind it and a coherent structural logic, rather than mere ideology. That seriousness is what makes the next section's counter-revolution a contest rather than a rout.
From the thesis Prebisch built the center-periphery frame — a world economy structured into an industrial center and a primary-producing periphery, with the structure itself reproducing the periphery's subordinate position — and from his post at the UN Economic Commission for Latin America (ECLA, or CEPAL in Spanish) he turned it into a policy program. That program was import-substitution industrialization (ISI): if specializing in primary commodities is a trap, the escape is to build domestic manufacturing behind tariff and quota walls, substituting home-made goods for imports and climbing the productivity ladder the terms of trade otherwise denied. ISI was the structuralist tradition's signature policy, and through the 1950s and 1960s it was the dominant development strategy across Latin America and much of the postwar developing world.
In its early phase the program even seemed to work. Behind protective tariffs, domestic industries that had not existed before — textiles, food processing, consumer durables, eventually steel and automobiles — were built, employment in manufacturing rose, and the larger Latin American economies posted respectable growth through the 1950s and into the 1960s. The trouble came at what theorists called the second stage. The easy substitutions, the simple consumer goods a protected market could absorb, ran out; the next round demanded capital goods and intermediate inputs that required scale, technology, and foreign exchange the inward-oriented economies were precisely structured not to earn. Protected from competition, the new industries had little reason to become efficient, and the home markets were too small to let them grow into world-class producers. The structuralist diagnosis of the problem was serious; whether its prescription could carry an economy past the second stage was the question the record would answer, and answer harshly.
The structuralist insight had a radical wing in dependency theory, and the wing split. Andre Gunder Frank pushed the argument to its sharpest form with the phrase "the development of underdevelopment": the periphery's poverty, on Frank's account, was not an earlier stage the rich countries had already passed through but an active product of the periphery's subordinate integration into the world capitalist economy. The center did not merely fail to develop the periphery; it developed itself by underdeveloping the periphery, extracting its surplus through the structure of trade and investment. The implication was near-revolutionary — a periphery could not develop within the world system, only by exit from it. Fernando Henrique Cardoso, writing with Enzo Faletto, drew a more conditional conclusion. Dependency, on their account, did not foreclose all development; it shaped a particular, distorted kind — "dependent development," growth that occurred but on terms set by the center and the local elites tied to it. The distance between Frank's exit-or-nothing and Cardoso's dependent-but-possible is the distance between dependency theory as a call to revolution and dependency theory as a structural account of how peripheral capitalism actually grows; Cardoso, who would later govern Brazil as a market-friendly president, embodied the gap. Whether the structuralist program's policy record justified its theory is the question section 16.4 takes up; the point here is that the theory was serious, and that import-substitution had a real intellectual argument behind it, not just a protectionist reflex.
The dependency school left a longer mark on the social sciences than on economics proper, and the asymmetry is itself instructive. Within economics the center-periphery framework was largely absorbed or set aside as the counter-revolution gathered force and the East Asian record cut against it; the discipline kept the terms-of-trade question as an empirical matter and dropped the structural-Marxist apparatus that surrounded it. But in sociology, political science, and the emerging field of world-systems analysis — Immanuel Wallerstein's account of a single capitalist world-economy stratified into core, periphery, and semi-periphery descends directly from this lineage — the dependency frame became foundational. That divergence marks a boundary this book respects: the structuralist insight about how the world economy distributes its gains is one thing, and the world-systems tradition that grew from it is another, treated where it belongs rather than re-narrated here.
The strongest serious case against it is structural, and it is Prebisch's.
The strongest serious case against free trade for poor countries is Prebisch's: comparative advantage in primary commodities can be a trap, not a gift, if the terms of trade drift against you. This is the structuralist for-voice the walkthrough mounts against the Ricardian gains-from-trade case — the same trade logic read as a mechanism of dependence.
By 1980 the structuralists had a problem their theory could not explain: the countries that had ignored their advice were the ones getting rich. Import-substitution, the program a generation of development economists had endorsed, had by the 1970s and 1980s produced a recognizable syndrome across much of Latin America and South Asia. Protected domestic industries, sheltered from competition behind tariff walls, stayed inefficient and small — building cars and appliances at multiples of world cost for a captive home market too small to capture scale economies. The protection generated rent-seeking: because an import license or a tariff exemption was worth more than any productive investment, the most profitable use of a firm's energy became lobbying the state for protection rather than cutting costs. Fiscal deficits and balance-of-payments crises followed, as governments subsidized the loss-making industries and the economies failed to export their way to the foreign exchange they needed. The gap between the theory's promise of catch-up and the program's actual result was wide, and it was visible.
The neoclassical reaction made that gap its central exhibit. Anne Krueger's 1974 paper named and modeled rent-seeking, showing that the social cost of protection was not just the standard deadweight loss but the additional resources squandered competing for the rents protection created. Jagdish Bhagwati extended the analysis of how trade restrictions generated directly-unproductive profit-seeking. And Deepak Lal, in The Poverty of "Development Economics" (1983), drew the sharpest conclusion: that there is no special development economics. The structuralists' whole premise — that poor economies need their own models because standard price theory does not apply to them — was, on Lal's account, exactly backwards. Get prices right, open the economy to trade and competition, and standard economics applies to poor countries as well as rich ones; the supposed structural peculiarities were mostly the distortions that bad policy had itself created. The counter-revolution did not merely criticize ISI; it questioned whether the field that produced ISI should exist as a separate field at all.
The era's load-bearing evidence was a comparison: East Asia versus Latin America. Through the same development decades, the outward-oriented economies of East Asia — Japan first, then South Korea, Taiwan, Singapore, Hong Kong — oriented their industries toward export markets and converged dramatically toward rich-world incomes, while the inward-oriented ISI economies of Latin America stalled and then, in the debt crisis of the 1980s, fell into a "lost decade" of contraction. The same decades, opposite outcomes. To the counter-revolution this was proof: openness beat inward orientation, the market beat the plan, and the East Asian miracle vindicated getting prices right. The trajectories themselves are the intellectual evidence in the debate, and they are worth reading directly — not as a history of the tigers (that history is Economic History chapter 17's) but as the exhibit the structuralist-versus-neoclassical argument turned on.
See the tigers converge and the ISI economies stall on the real trajectories. The East Asian exporters (South Korea, Singapore, Japan) climb toward rich-world incomes through the same decades the inward-oriented Latin American economies (Argentina, Brazil, Mexico, Chile) stall and then fall into the 1980s debt crisis. Read the divergence as the intellectual evidence in the structuralist-versus-neoclassical debate — and keep the contested reading in view: was it the free market that did it, or a capable developmental state?
Figure 16.3 (interactive). The East-Asia-versus-Latin-America divergence on the GDP-per-capita map. Scrub the timeline; click a country for its annotation panel. The map is the source of truth; this is a contextualized viewport into it. The exhibit teaches lineage, not just data — the divergence is real, but its reading is contested.
That contested reading is where the era's other lineage was born, and the history-of-economic-thought account makes its home here. The World Bank's own 1993 study, The East Asian Miracle, read the record in a deliberately "market-friendly" way: the tigers succeeded because they kept prices roughly right, invested in education, and stayed export-oriented, with selective state intervention playing at most a supporting role. A rival reading insisted the state had been load-bearing. Chalmers Johnson's MITI and the Japanese Miracle (1982) coined the developmental state to describe a state that deliberately steered industrial transformation — targeting credit to favored sectors, disciplining firms through export performance, and tying protection to results rather than handing it out unconditionally. Alice Amsden's Asia's Next Giant (1989) carried the reading to South Korea, arguing that Korea grew precisely by "getting prices wrong" on purpose — subsidizing and protecting strategically while imposing reciprocal performance conditions that ISI's Latin American version never imposed. Robert Wade's Governing the Market (1990) made the case for Taiwan, and Peter Evans's Embedded Autonomy (1995) drew the comparative lesson: the developmental state worked where the bureaucracy was both insulated enough to resist capture and connected enough to coordinate with industry. The contrast with ISI was not openness versus protection — the tigers protected too — but disciplined, performance-conditioned intervention versus the unconditional, rent-generating kind. This developmental-state lineage is the seed of a debate whose current standing belongs to chapter 17 and to the industrial-policy walkthroughs; here it is named as the lineage's other reading of the very evidence the counter-revolution claimed.
The counter-revolution's prescriptions were eventually codified. In 1989 John Williamson enumerated the package of reforms that the Washington-based institutions — the IMF, the World Bank, the US Treasury — had converged on for developing economies, and called it the Washington Consensus: fiscal discipline, tax reform, trade liberalization, openness to foreign investment, privatization, deregulation, secure property rights, and the rest. As a description of a real convergence the list was accurate. As a universal prescription it aged badly. The structural-adjustment programs that carried it into practice through the 1980s and 1990s too often delivered austerity and recession rather than the promised catch-up; the "Washington Consensus" became, within a decade, a near-pejorative, shorthand for a one-size-fits-all market orthodoxy that prescribed more than its evidence supported. Joseph Stiglitz, from inside the World Bank, became its most prominent critic, arguing that the Consensus had ignored market failures, sequencing, and institutions, and had imposed a template indifferent to local conditions.
The sharpest evidence against the Consensus as a universal template came from sub-Saharan Africa, where the structural-adjustment programs of the 1980s and 1990s were applied most thoroughly and produced the least growth. Currency devaluation, trade liberalization, the dismantling of marketing boards, and deep cuts to state spending were imposed as the price of debt relief and new lending, and across much of the continent they delivered stagnation, deindustrialization, and collapsing public health and education systems rather than the promised recovery. By the late 1990s even the Bretton Woods institutions had begun to retreat from the harder version of the program, folding poverty-reduction and institution-building language into what had been a purely market-liberalizing agenda. The retreat was the clearest admission that the over-correction was real: a doctrine confident enough to be imposed as a condition of survival turned out to rest on evidence that did not travel from the East Asian cases that inspired it to the African ones where it was tested.
The honest call here is a settled one, and it cuts both ways without splitting the difference. The neoclassical counter-revolution corrected a real over-reach: ISI's record by the 1980s genuinely was poor, the East Asian outward orientation genuinely did outperform inward-oriented import-substitution, and Krueger and Lal were right that protection breeds rent-seeking and that much of what the structuralists called structure was self-inflicted distortion. That is a call to take squarely, not a both-sides to hedge. But the counter-revolution over-corrected: the Washington Consensus prescribed a uniform market template that exceeded what the East Asian evidence actually supported, since the tigers that vindicated openness had also relied on exactly the disciplined state intervention the Consensus forbade. The correction was real and the over-correction was real. Both are settled; the genuinely open question — whether the state's developmental role was the decisive ingredient or a replaceable one — is what the lineage carries forward to the modern industrial-policy debate.
The exhibit you just read has two readings, and this is the second.
Japan and Korea didn't just open up — the developmental-state reading (Johnson, Amsden, Wade) says the state was load-bearing: directed credit, export discipline, protection tied to performance. That is the lineage that contests the World Bank's market-friendly reading of the same East Asian record — the developmental state's intellectual home.
This is where state-formation meets development economics.
The developmental state is where state-formation meets development economics — the state as the deliberate agent of structural transformation, not just the keeper of order. Johnson's Japan, Amsden's Korea, Wade's Taiwan, Evans's comparative cases are the lineage's evidence that a capable, insulated-yet-embedded bureaucracy can steer a late industrialization.
Every model so far has measured development in dollars per head. Amartya Sen asked a prior question: dollars for what? A wealthy man in the middle of a famine and a poor man with enough to eat — which one is developed? Income tells you nothing about the famine; it tells you only about the dollars. Sen's contribution was not a new growth theory or a new method but a relocation of the question itself — a change in what development economics is for — and it has reshaped how the rest of the field's questions get read.
The relocation runs through the capabilities approach, which Sen began stating in his 1979 lecture "Equality of What?" and developed across the following two decades into Development as Freedom (1999). Development, on this account, is the expansion of substantive freedoms — the real opportunities a person has to live a life they have reason to value. The unit is the capability: the freedom to achieve various functionings, where a functioning is a state of being or doing a person actually achieves — being adequately nourished, being literate, being healthy, taking part in the life of the community, appearing in public without shame. The capability is the freedom to achieve such functionings; a person's well-being is the set of capabilities open to them, and development is the widening of that set. The distinction between capability and functioning is load-bearing: two people may achieve the same functioning — both adequately fed — but one because food is plentiful and chosen, the other because fasting is the only food security available. The first has the capability and chooses to be fed; the second lacks the capability and is fed only by necessity. Income cannot tell them apart.
The force of the move is in what it rules out as the right metric. Income fails because it is a means, not an end — the famine victim's problem is not a shortage of dollars in the abstract but a collapse of the entitlements that convert dollars into food, and the same income buys radically different capability sets for a healthy person and a chronically ill one, a literate person and an illiterate one. Utility, the welfarist's metric, fails because it adapts: the chronically deprived learn to want little, so a contented but oppressed person can register high utility while their capabilities are crushed, and a metric that calls adaptive resignation "welfare" cannot be the measure of development. Even Rawls's primary goods — the all-purpose means a just society distributes — fail, Sen argued, because people differ in how efficiently they convert goods into functionings: a disabled person needs more resources to achieve the same mobility, so equal primary goods leave unequal capabilities. Capabilities are what survive these objections, because they measure not what a person has or feels but what a person is actually free to do and be.
The redefinition got operationalized, and that is what gave it staying power. Working with Mahbub ul Haq, Sen helped design the Human Development Index (HDI), introduced in the first Human Development Report in 1990. The HDI is a deliberately crude composite — income, life expectancy, and education, combined into a single number — and its crudeness was the point: a measure that put a capability-informed alternative to GDP into annual practice, ranking countries by a richer notion of development than income alone, and doing so visibly enough that a country could no longer hide a development failure behind a respectable GDP. The HDI was one particular operationalization of Sen's idea, not the idea itself; the deeper claim is that what counts as development is a choice of metric, and that choosing income silently builds in an answer to a question that should be argued.
The capabilities frame was not Sen's only contribution to how development is understood, and his earlier work on famines is what gave the frame its empirical bite. In Poverty and Famines (1981) Sen showed that the great famines of the twentieth century — Bengal in 1943, the Sahel, Bangladesh — typically occurred without any collapse in the aggregate food supply: people starved not because there was too little food in the country but because they had lost the entitlements — the wages, the crops, the exchange relations — that let them command it. Famine was a failure of distribution and access, not of production, and a development metric that tracked only aggregate output would have registered no problem at all while millions died. The entitlement analysis is the capabilities approach in its starkest form: what matters is not how much a society produces but what its members are actually able to obtain and to do, and the two can come catastrophically apart. That insight, more than any index, is why the redefinition of the development objective stuck.
Shift the weight you place on income versus health versus education, and watch a fixed set of countries re-rank. A country that ranks high on income alone can drop sharply once health and education enter the score — which is Sen's whole point: "what counts as development" is a choice of metric, not a fact you read off GDP. The "HDI default" button snaps the weights to the equal-thirds the 1990 index used, one particular operationalization of that choice.
Figure 16.4 (interactive). A small fixed country set scored on a weighted composite of income, health, and education, and ranked. Weights re-normalize to sum to one. Drag a country high on income toward a capability weighting and watch it fall — the re-rank is Sen's claim made drivable. Indicators are stylized for illustration, not a sourced dataset.
A country can be rich and short-lived, or poorer and long-lived and well-schooled. Which is "more developed" is not a fact — it is a decision about what to weight. Move the weights and the leaderboard reshuffles, which is exactly why Sen says income silently smuggles in an answer to a question that should be argued. A still-broader measure would count what even the HDI misses — the unpaid care work that sustains every economy and that no national income account records.
The HDI's construction and the broader facts of development are in Economics chapter 20.1. This widget computes a stylized weighted composite to make the metric-dependence of rankings drivable, not the official HDI formula.
The relocation reaches forward in two directions worth marking. First, it reframes the question the randomized-trial turn is about to raise. When that generation asks "what works?", Sen's redefinition forces the prior question back into view: works for what end? An intervention that raises income but not capabilities, or that improves a measured functioning while narrowing the freedoms behind it, has not straightforwardly "worked," and the capabilities frame is what keeps that distinction alive. Second, the capabilities approach opened a door the feminist economists walked through. If development is about what people are free to do and be, then the unpaid care work that sustains every household and economy — overwhelmingly done by women, counted in no national income account — is exactly the kind of contribution income measures miss. Nancy Folbre's critique of how economies systematically undercount the work of social reproduction, and Marilyn Waring's earlier demonstration of what the national accounts leave out, are the capabilities frame widened to the household. That Sen-to-Folbre handoff is named here as a thread; the full beyond-GDP measurement movement, the broader Stiglitz-Sen-Fitoussi line, belongs to chapter 17. Sen took the Nobel in 1998. Where he sits relative to the lineage is on the Sen node of the timeline.
The first institutionalized answer was built here.
Sen's capabilities and the HDI were the first institutionalized answer to "is GDP misleading?" — a measure that put income, health, and education together because development is not dollars. And the Sen-to-Folbre handoff widens it further, to the unpaid care work GDP never counts.
Sen reframed what the question is even asking.
Sen's capabilities reframe the poverty question: poverty is a deprivation of freedoms — of the real opportunities to be nourished, healthy, schooled, heard — not just a low income to be raised. That shifts "distribution or growth?" onto a different axis, since both growth and redistribution matter only insofar as they expand what people can actually do and be.
Sen partly dissolves the question.
Sen reframes the democracy-and-growth question: development is freedom, so the political freedoms democracy secures are not merely a means to growth but part of what development consists in. On that reading the question partly dissolves — "does democracy cause growth?" assumes growth is the end, which is exactly what Sen denies.
The capabilities frame and its feminist extension are part of the answer.
A coherent left economics leans on the capabilities frame and its feminist-economics extension (Folbre) — what GDP and income miss: the unpaid care work and the substantive freedoms that markets neither price nor produce. Sen and the social-reproduction critique are one of the standing-tools strands the walkthrough assembles.
For forty years development economists argued about what makes nations grow. Then a generation decided the argument was unwinnable on the available evidence — and started measuring whether a deworming pill keeps kids in school. The shift was not modest. It was a wholesale change in what the field counted as knowledge, and it earned a Nobel Prize and reorganized how billions of dollars of aid get spent. It also, by design, walked away from the question Lewis had asked.
The canonical result is the deworming pill, and it shows what the new method bought. In the mid-1990s Michael Kremer ran field experiments in western Kenya, randomly assigning schools to receive cheap mass deworming treatment and comparing them against schools that did not. The finding was striking: deworming raised school attendance more per dollar than almost any conventional education intervention — not by making children smarter but by keeping the sick ones in class. The result was specific, credibly identified, and cheap to scale, and grand growth theory could never have produced it — not because growth theorists were careless but because the question "does deworming raise attendance?" is simply not the kind of question a theory of aggregate growth is built to answer. It is the kind a randomized experiment answers cleanly. That contrast — a small, answerable, credibly-measured question delivering a real result where a large, unanswerable one delivered decades of argument — is the whole case for the turn.
The core move was to stop building grand theories of why nations grow and instead run randomized controlled trials on specific development interventions — microcredit, conditional cash transfers, bed nets, fertilizer subsidies, remedial tutoring, deworming — evaluating each on credible causal identification rather than on theory. Abhijit Banerjee and Esther Duflo, with Kremer, built the apparatus and the institution: the Poverty Action Lab, founded at MIT in 2003 and later J-PAL, became the hub of a movement that ran hundreds of field experiments across the developing world, and their book Poor Economics (2011) made the case to a wide audience — the argument that the right way to fight poverty is not to settle the big ideological questions (aid versus markets, the role of the state) but to find out, intervention by careful intervention, what actually works. The 2019 Nobel went jointly to Banerjee, Duflo, and Kremer "for their experimental approach to alleviating global poverty," and Duflo, at forty-six, became the youngest economics laureate and the second woman to win.
The microcredit reckoning shows the method working at its sharpest, against the field's own enthusiasms. Through the 2000s, small-loan programs in the mold of Muhammad Yunus's Grameen Bank were the development world's favorite story — tiny loans to poor entrepreneurs, especially women, were held to lift households out of poverty and won Yunus the 2006 Nobel Peace Prize. The randomized evaluations, run across India, the Philippines, Morocco, Mexico, and elsewhere, told a more sober story: microcredit did expand borrowers' choices and let some businesses grow, but the transformative poverty-reduction effect the movement had promised simply did not appear in the data, with no reliable gains in average household income, health, or schooling. The result did not refute microcredit so much as right-size it — a useful financial product, not a cure for poverty — and it is exactly the kind of correction the older method could not perform, because the case for microcredit had rested on inspiring anecdotes and theory rather than on the counterfactual the trials supplied. Telling a genuine effect from a compelling story is what the machine is for, and microcredit is the case where it told them apart most consequentially.
What licenses the method is the logic of random assignment, and it is the same logic the credibility revolution carried across all of empirical economics. The problem with observational comparison is selection. If you compare villages that adopted a program against villages that did not, the two groups differ not just in the program but in whatever made some villages adopt and others not — better governance, more enterprising leaders, easier access — and that difference, not the program, may drive any outcome gap. Random assignment breaks the link: when a coin flip decides which villages get the program, the treatment and control groups are, on average, identical in everything except the treatment, so the difference in their outcomes is the program's causal effect and nothing else. The development RCT is the field application of the credibility revolution that chapter 12 narrates as the discipline's reigning evidence standard — the same identification-first commitment that drove the natural-experiment and instrumental-variable work in labor and elsewhere, carried into the field where the experiment could be designed rather than found. (Its predecessor rung, the time-series and natural-experiment turn, is chapter 11's.) The formal estimation apparatus — the local-average-treatment-effect machinery, the potential-outcomes framework — lives in Economics chapter 10.6 and chapter 20.6; here the point is the move's place in the lineage, not its mathematics.
The turn reorganized practice as well as theory, and the reorganization is part of why it earned the Nobel rather than remaining an academic fashion. J-PAL and its sister networks turned the trial into an institution: a pipeline that took a candidate intervention, evaluated it under randomization, and, if it survived, pushed for it to be scaled by governments and donors — deworming campaigns reaching hundreds of millions of children on the strength of the cost-effectiveness Kremer's studies established, conditional-cash-transfer programs spreading from Mexico's Progresa across Latin America and beyond after trials confirmed they raised schooling and health, chlorine dispensers and free bed nets adopted because the trials showed that charging even a token price collapsed uptake. The effective-altruism movement and evaluators like GiveWell built an entire philanthropic apparatus on the premise that the right way to give is to fund the interventions the trials had shown to work per dollar. This is a real and unusual achievement for a social science — a body of findings that changed where money goes and what governments do, grounded in evidence that survives scrutiny because the comparison was clean. It is also, precisely in its strength, the source of the limitation the next section turns on: every item in this catalogue is an intervention, scalable across villages and verifiable in a trial, and not one of them is the structural transformation that turns a poor country into a rich one.
It is worth being precise about what the turn was reacting against, because the reaction defines it. The grand theories had not failed for want of effort; they had failed for want of a way to be checked. A planner could believe in the big push or in dependency or in the developmental state and find historical episodes that seemed to confirm each, because the aggregate growth record is a single, non-repeatable run of history that no theory can be cleanly tested against. The randomized generation's deepest claim was epistemological rather than substantive: that the reason development economics had argued in circles for forty years was that it had been asking questions its evidence could not adjudicate, and that intellectual honesty required scaling the question down to the size at which an answer could actually be earned. Whether that was humility or surrender is the disagreement the lineage closes on; that it was a considered epistemological choice, and not mere technical fashion, is what gives the debate its weight.
Estimate a program's effect two ways. Set the mode to observational — comparing villages that adopted against villages that didn't — and drag the selection-strength slider up: the naive estimate drifts away from the true effect, because the adopting villages differ systematically on a confounder. Now flip to randomized assignment and drag the same slider: the estimate collapses onto the true effect no matter how strong the selection. That is why random assignment identifies what observation cannot — and the contrast is what a static sentence can only assert.
Figure 16.5 (interactive). The estimated treatment effect against selection strength. Under observation the estimate is biased by selection; under randomization it tracks the true effect regardless. Drag the selection slider in each mode. The estimate = true effect + (selection × confounder) under observation; = true effect + sampling noise under randomization.
When villages choose for themselves whether to take a program, the ones that choose it are different to begin with — and that difference contaminates the comparison, the more so the stronger the selection. A coin flip removes the choosing: treatment and control end up alike in everything but the program, so the outcome gap is the program and nothing else. That is the whole machine — and notice what it is silent on. It tells you cleanly whether the pill helps. It says nothing about what makes Kenya grow.
The local-average-treatment-effect machinery, the potential-outcomes framework, and the external-validity econometrics are in Economics chapter 10.6 and chapter 20.7. This widget computes a stylized bias to make the identification logic drivable, not the estimator.
What the turn delivered is real and what it set aside is real, and the second is what the standing debate turns on. On the delivery side: a genuine, cumulating body of what-works knowledge — that deworming beats most education spending per dollar, that the headline microcredit promise was oversold, that conditional cash transfers reliably raise schooling and health, that small frictions and behavioral nudges matter more than the planners assumed — and policy adoption at scale, with governments and donors redirecting billions toward interventions that survived a trial and away from those that did not. On the set-aside side: the aggregate growth question, by design. An RCT measures an intervention; it cannot measure a development trajectory. You can randomize bed nets across villages; you cannot randomize industrial policy across countries, or run a controlled trial of what turned South Korea from a Lewis economy into a rich one. The turn's discipline about evidence was bought by narrowing the question to the size at which evidence could be clean — and whether that narrowing was a gain in honesty or a loss of nerve is the standing debate that closes the lineage.
Decades after Lewis, the field changed the subject.
And the modern field largely stopped trying to answer the rich/poor question at the level of whole economies — turning instead to RCTs on what works, intervention by intervention. The clean evidence came at the cost of the original question, which is exactly the trade the walkthrough's closing positions weigh.
The RCT turn sidesteps the framing dispute the question stages.
The Sachs-versus-Easterly framing dispute — big-push aid to escape the trap versus let-markets-and-incentives-work — is the structural pole against the skeptic. The RCT turn sidestepped both by refusing to theorize the whole economy and measuring interventions instead: not "why poor?" but "does this specific thing help?"
The development RCT is its maximum-clean instance.
The development RCT is the credibility revolution at its maximum-clean instance — the experiment designed rather than found. The same machine that made "does this pill help?" answerable made "what makes a nation grow?" the question it declines. The discipline-wide arc is chapter 12's; this is its development stop.
The pendulum that swung from Lewis's grand model to Duflo's randomized pill bought one thing and sold another: it bought credibility and sold ambition. Whether that was a good trade is the question the field has not settled, and it is worth stating the standing debate as the sharp confrontation it actually is. On one side stands the randomized-trial economist who refuses to theorize about growth because the evidence for any such theory is not credible — who would rather know one thing cleanly than guess at everything. On the other stands the structural economist who refuses to abandon the growth question because that, not the deworming pill, is what development actually is — who would rather chase the right question with imperfect tools than answer a smaller one perfectly. Both refusals are principled. The field has not adjudicated between them.
The credibility critique's gains are real and should be granted in full. The randomized turn ended a long era of grand-theory hand-waving in which development economists argued about the engines of growth with no way to settle the argument, and it replaced that with a standard of evidence the field had never had — internal validity, credible causal identification, results that survive scrutiny because the comparison was clean. The body of what-works knowledge it produced is genuine and policy-relevant in a way the grand theories never managed; on the merits of evidence, the bar rose, and it rose permanently. None of that is in dispute.
The structural critique's force is equally real, and it comes from serious economists. Angus Deaton, in his 2010 critique and elsewhere, argued that the RCT movement had elevated internal validity at the expense of external validity — the question of whether a result established in one Kenyan district holds in another district, another country, another scale — and that randomization, for all its rigor at home, gives no special warrant for travel. He coined the charge that the movement suffered a "tyranny of the average treatment effect," reporting a mean impact that may describe no actual person and conceal the heterogeneity that policy most needs to understand. Martin Ravallion pressed the external-validity and scaling problems from the World Bank side. Lant Pritchett put the sharpest version: what he called "kinky development," the objection that the RCT movement measures interventions while the development question is about growth — that randomizing bed nets, however cleanly, tells you nothing about the structural transformation that turns a poor country into a rich one, and that a generation of the field's best minds had answered a question adjacent to the one that matters because it was the question they could answer. Dani Rodrik occupies the chastened middle: the structuralist tradition's modern heir, he abandoned the grand theory but kept the question, proposing "growth diagnostics" — a country-by-country search for the binding constraint on growth — as a way to be disciplined about the aggregate question without pretending one model fits all.
There is another response to the growth question the RCTs decline, and it is the territory of chapter 15, not this one. The institutions-as-deep-cause answer — the argument associated with Daron Acemoglu and James Robinson that the deep determinant of whether a country is rich or poor is whether its institutions are inclusive or extractive, evidenced by the settler-mortality instrument and argued in Why Nations Fail — is one of the live alternatives that kept the growth question open precisely when the RCT turn set it aside. It belongs to the institutional tradition; here it is named, not developed, as one branch of the same growth question that the credible-measurement turn chose not to answer.
So the chapter's call is a call about what is settled and what is not. What is settled: the credibility revolution genuinely and permanently raised the field's standard for what counts as evidence, and that gain will not be given back. Also settled, from earlier sections: ISI's record was poor, the Washington Consensus over-prescribed. What is not settled, and what the chapter declines to settle because the field has not: whether "why do nations grow?" is answerable by the new methods at all. The credible-measurement turn bought its internal validity with scope, and whether that trade was wisdom or evasion depends on a judgment about whether the growth question is answerable that no one yet has the evidence to make. The honest position is to hold both poles at full strength and name the dispute open — not "the RCTs won," not "the structuralists were right all along," and not the mushy middle that says the truth lies somewhere between. The structural-transformation tradition and the institutions tradition stay live because the RCTs declined their question, and that is not a failure of the field to converge but an accurate reflection of a question the available methods cannot yet close.
The stakes of leaving the question open are not merely academic, and they are sharpest where the lineage began. The countries that most need a theory of how a whole economy develops — the low-income economies of sub-Saharan Africa and South Asia that have not had an East Asian take-off — are precisely the ones for whom the catalogue of credibly-tested interventions, real as it is, adds up to less than a development strategy. Knowing that deworming beats most education spending per dollar tells a finance minister something genuinely useful; it does not tell her how to turn an agrarian economy into an industrial one, which is the transformation Lewis set out to model and the one the structural and institutional traditions still claim as their subject. The honest reckoning is that the field traded a question it could not answer credibly for a set of questions it could, and that the trade was both a real gain in rigor and a real narrowing of ambition. Whether the original question is answerable at all by methods like these, or whether it requires a different kind of knowledge the discipline has not yet built, is the live disagreement that keeps development economics from settling into a single method — and the reason its history is a pendulum rather than a ladder.
It is worth being plain about what the intellectual-history timeline does and does not yet carry of this lineage. The development cluster on the graph holds Prebisch, Sen, Banerjee, and the development-economics school node, with Acemoglu adjacent as the institutions bridge. It does not yet carry Lewis, Rostow, Rosenstein-Rodan, Nurkse, Hirschman, Singer, Duflo, Kremer, or the developmental-state cluster of Johnson, Amsden, Wade, and Evans — the founding figures and the modern field's other half. The relational view below is the development school filtered out of the larger graph; read against the gap just named, it shows the cluster the lineage has so far, with the founding nodes still to be added.
For the fuller relational view, the development school on the intellectual-history timeline shows the cluster's present members and their live edges; the founding figures named above are the lineage's not-yet-mapped nodes, and seeing the two poles of the pendulum — structural theory and credible measurement — as positions in one connected field is the relational counterpart to the argument this chapter has walked. The graph shows the structure; the call about whether the question is answerable stays in the prose, where the chapter takes it openly and leaves it open.
Where the lineage's modern heir lands.
Rodrik is the chastened structuralist heir — growth diagnostics, not grand theory: a country-by-country search for the binding constraint, keeping the developmental-state question alive without pretending one model fits all. He is where the lineage's structural pole stands after the credibility turn.
The lineage's live question, stated at the close.
And the lineage's live question is whether the developmental state's agency is replicable or contingent — whether a poor country today can assemble the insulated-yet-embedded bureaucracy the tigers had, or whether that capacity was a product of particular histories that cannot be engineered. That is the unresolved end of the state-formation thread.
Lewis, "Economic Development with Unlimited Supplies of Labour" (1954). Rostow, The Stages of Economic Growth: A Non-Communist Manifesto (1960). Rosenstein-Rodan, "Problems of Industrialisation of Eastern and South-Eastern Europe" (1943). Nurkse, Problems of Capital Formation in Underdeveloped Countries (1953). Hirschman, The Strategy of Economic Development (1958). Prebisch, The Economic Development of Latin America and Its Principal Problems (1950); Singer, "The Distribution of Gains between Investing and Borrowing Countries" (1950). Frank, "The Development of Underdevelopment" (1966). Cardoso and Faletto, Dependency and Development in Latin America (1979). Krueger, "The Political Economy of the Rent-Seeking Society" (1974). Bhagwati, "Directly Unproductive, Profit-Seeking Activities" (1982). Lal, The Poverty of "Development Economics" (1983). Johnson, MITI and the Japanese Miracle (1982). Amsden, Asia's Next Giant (1989). Wade, Governing the Market (1990). Evans, Embedded Autonomy (1995). World Bank, The East Asian Miracle (1993). Williamson, "What Washington Means by Policy Reform" (1990). Sen, "Equality of What?" (1979); Development as Freedom (1999). UNDP, Human Development Report (1990). Folbre, The Invisible Heart (2001). Banerjee and Duflo, Poor Economics (2011). Kremer and Miguel, "Worms" (2004). Deaton, "Instruments, Randomization, and Learning about Development" (2010). Ravallion, "Should the Randomistas Rule?" (2009). Pritchett, "It Pays to Be Ignorant" and the "kinky development" critique. Rodrik, One Economics, Many Recipes (2007).