The chapter opens not with a definition but with six testimonies. They are worth reading closely, because the definition of development the chapter eventually gives is assembled out of exactly the things they mention.

Read together, the six say something a single income statistic cannot. The Ugandan woman is describing the loss of voice before she gets to the loss of food. The Moldovan woman is describing shame. The Ethiopian is describing vulnerability. Only the Brazilian mentions income directly, and even there it is bundled with medicine and clothing. The slide draws the conclusion in one line:

That single sentence is the chapter's thesis. If poverty has several dimensions, then a development policy that raises income while leaving the other dimensions untouched has not, by this definition, succeeded.

Three scopes, not three opinions

Development economics is not a political stance within economics; it is a different scope. The slide sets it beside two neighbours, and the comparison is examinable as a comparison.

Key features of developing economies

These four features are the reason the neoclassical apparatus cannot simply be transplanted. Each one is a textbook assumption being switched off.

The slide's summary line: development economics studies the economic, social, and institutional mechanisms that raise living standards at scale.

The important role of values

Development is not value-free — judgements of "improvement" are inherently normative. Policy involves trade-offs between efficiency, equity, growth, stability and sustainability, and the slide is explicit that the focus on poverty and freedom rests on normative commitments to dignity and justice. This is not a disclaimer; it is a claim that the subject matter cannot be described without choosing what counts as better.

Economies as social systems

Four claims, each of which rules out a purely economic answer:

• Economies are embedded in institutions, culture, history, and politics.
• Development policy depends as much on institutional variables as on capital and labour.
• Policies that work in rich countries often fail in developing contexts.
• A holistic, interdisciplinary approach is therefore required.

The traditional economic measure

National development = growth in GNI or income per capita, with income as the main driver of utility. Chapter 2 will show you how this is measured and why the measurement is harder than it looks. The point here is narrower: this is the definition the rest of section 1.3 is arguing against.

The new view: Sen's capability approach 可行能力

Amartya Sen's reformulation replaces income with freedom as the thing to be maximised. Development is the expansion of human capabilities and substantive freedoms. Four terms carry the whole argument:

Some key capabilities

The slide splits them into "beings" (states of existence) and "doings". Each list is a checklist rather than a hierarchy.

The three core values of development

This is the most examinable list in the chapter. Three values, each with a short definition and a stated goal.

Income and happiness: Figure 1.2

The slide plots happiness (vertical) against income per head in US$ per year (horizontal, running to about $35,000), one dot per country, using Layard's data. Three numbered conclusions are printed on it:

The scatter is a steep rise at low income and a near-flat cloud at high income. Ukraine sits at the bottom of the vertical scale, in the low 30s; the high-happiness cluster (roughly 80–100) contains countries spread across a very wide income range, which is the second conclusion made visible — the same happiness level is reached at $10,000 and at $35,000.

The central role of women

Women's empowerment is both a core goal and a key driver of development — it improves child health, education, and long-run human capital. The slide states it as a two-way relationship, and the distinction matters: even if you cared nothing about equity between men and women, the growth evidence would still make female education one of the highest-return investments available.

The three objectives of development

1. Increase the availability of life-sustaining goods Food, shelter, health, protection
2. Raise overall levels of living Higher incomes, more jobs, better education, greater attention to cultural and human values
3. Expand the range of economic and social choices By freeing people from servitude and dependence

These three objectives are the three core values restated in operational form — the same three ideas, said once as values and once as aims.

Eight goals adopted by the United Nations in 2000, with a 2015 deadline. They are worth knowing as a list, because the SDGs that replaced them are an extension of the same structure.

Table 1.1 — the goals and their targets

The full table runs to two slides. The targets are the measurable half of each goal:

From the MDGs to the SDGs

The MDG era ended in 2015. In September of that year UN leaders adopted the 2030 Agenda, launching 17 Sustainable Development Goals to be achieved over the following fifteen years. The SDGs are broader than the MDGs — they apply to all countries rather than only to developing ones, and they add goals on inequality, energy, cities, and climate that the MDGs never covered.

Conclusions

The chapter closes on three points: the importance of development economics; the inclusion of non-economic variables in designing development strategies; and the effort to achieve the MDGs. The slide's closing quotation is "…One future — or none at all." The instructor also assigned the documentary Why Poverty?, Episode 1: Poor Us, which is the same argument as section 1.1 in film form.

Section A tests the definitions you would be asked to write out. Section B tests the distinctions that are easy to blur — functionings against capabilities, the three values against the three objectives. Section C asks for the extended answers.

The World Bank's scheme is the one used in practice: countries are ranked on GNI per capita and sorted into four bands. The thresholds are revised every year, so the slide's 2025 numbers are the ones to quote.

The slide also shows the alternative grouping the course uses when geography matters more than income: by region. The two groupings disagree on purpose. A country can be upper-middle-income and in Sub-Saharan Africa, or low-income and in East Asia, and which grouping is useful depends on the question — income explains what a government can afford, region explains what its neighbours are doing.

GNI and GDP

Two accounting identities that get confused constantly. The difference is who is being counted, not what is being counted.

The expenditure approach to GDP

Y = GDP in dollars · C = consumption · I = investment · G = government purchases · NX = net exports = exports − imports

This is the national income accounting identity. It is useful because it breaks GDP into components according to their purpose: goods and services can be consumed, invested by the private sector, bought by the government, or shipped abroad. NX can be positive or negative — a country that exports more than it imports has positive net exports; the United States, which imports more than it exports, has negative net exports and correspondingly a trade deficit.

Measuring changes over time

The slide's own illustration: US nominal GDP was $14.4 trillion in 2008 and only $7.4 trillion in 1995. Did prices change? Did quantities change? Some of both — and the whole point of real GDP is to answer the question that the nominal figure leaves open. The instructor's note frames it as the familiar (total revenue) = price × quantity: nominal GDP is the revenue line, real GDP holds the price term still.

Comparing incomes across countries needs a conversion factor, and the obvious one — the market exchange rate — answers the wrong question. It tells you what a dollar buys in currency markets; what you want to know is what it buys in the local market. In poor countries, non-traded goods and services (a haircut, a meal, a bus ride, housing) are far cheaper than they are in rich ones, so converting at the market rate systematically understates real incomes in poor countries.

What the difference looks like in the data

The instructor's table (slide 10) puts the two conversion methods side by side. The pattern is systematic: the poorer the country, the larger the upward revision from switching to PPP.

The ratio column is the story. For high-income countries PPP is barely different from the market rate — the United States is 1.03×, because it is the base. For low-income countries the revision is around two-and-a-half to three times, and the largest ratios in the table belong to India (3.88×), Egypt (3.77×) and Pakistan (3.69×). The low-income group average rises from $775 to $2,127 — a country that looks desperately poor at market rates looks merely poor at PPP.

The HDI is Sen's capability approach made operational: three dimensions — health, knowledge, and a decent standard of living — combined into one number between 0 and 1. The classification bands are:

The slide also stresses that the HDI can be calculated for groups and regions within a country, not just for countries — it varies among groups, across regions, and between rural and urban areas. That matters: the whole point of a holistic measure is that a national average can hide the thing you care about.

The old HDI, worked: Bangladesh in 2007

This is the slide's own example, with every intermediate number. The old HDI builds a sub-index for each dimension and then takes the arithmetic mean.

The goalposts are visible inside the formulas: life expectancy runs from 25 to 85 years, literacy and enrolment from 0 to 100%, and income from ln(100) to ln(40,000). The income index is logarithmic precisely because of the diminishing-returns result from Figure 1.2 in Chapter 1.

What is new in the New HDI

The UNDP rebuilt the index in November 2010. There are two headline changes and several smaller ones.

The smaller changes are just as examinable:

The effect of the whole package is that the New HDI can identify not only whether a country is developing, but also whether various significant groups within that country are participating in the development.

The New HDI, worked: China in 2010

Discussion of the HDI

Table 2.4 — HDI and its components, selected countries (2018)

The last column is the most interesting one. It is the country's GNI rank minus its HDI rank, so a positive number means the country does better on human development than its income alone would predict. Turkey is +43 and Bangladesh +20; Egypt is −80 and Brazil −40. The HDI's own advantage statement — "a country can do much better than might be expected at a low level of income" — is exactly what those positive numbers are showing.

The extremes, as of the instructor's 2026 update

The world average in 2023 is 0.72. Burundi and South Sudan are tied at rank 180 as printed on the slide. Note that Norway and Switzerland share 0.970 and are separated only by rounding.

The section's subtitle is the point: there are shared characteristics, but the diversity within them is enormous. The slides develop each characteristic with data, and the data is where the examinable detail lives.

Figure 2.4 — shares of global income, 2024

Nominal GDP, current US dollars. Seven shares, which sum to 100%:

Table 2.6 — the 12 most and least populated countries, 2024

Table 2.10 — the urban population, 2024

The developed/developing split and the regional breakdown are two separate partitions, so the regional rows do not sum to the world total. The check that does hold: 1,259 + 6,901 = 8,160.

Table 2.11 — employment by sector, 2004–2008 (%)

The single most useful table for the structural-change argument in Chapter 3: it shows agriculture absorbing most of the labour force in poor countries while contributing a small share of GDP — the gap between the two columns is the productivity difference Lewis's model is built on.

M = male, F = female, GDP = share of GDP (2008). Source: World Bank, World Development Indicators 2010, tabs. 2.3 and 4.2. Note on the slide: Ethiopian agricultural employment reflects limited coverage.

Table 2.8 — crude birth rates around the world, 2022

Crude birth rate 粗出生率 = births per 1,000 population per year. The slide lists roughly 150 countries; what matters for revision is the shape of the distribution, and where China and the other large developing countries sit.

The pattern is the demographic transition: the highest rates are concentrated in Sub-Saharan Africa, the lowest in Europe and East Asia. China, Thailand and Brazil now sit in the same band as the United States and the United Kingdom.

Table 2.7 — primary school enrolment and pupil-teacher ratios, 2022

Four cells in the printed table were not legible on the slide render and are marked "—" rather than guessed. The legible ratios run 37 → 28 → 14 for the income groups and 20 → 22 → 33 → 36 for the regions, so the missing values sit inside that band.

Figure 2.4 — under-5 mortality rates, 1990 and 2023

Deaths per 1,000 live births. Two of the eight bar labels and the high-income 1990 value were not legible on the slide. The direction is unambiguous: the largest absolute fall is in low-income countries (128 → 38), which is the same distance travelled as lower-middle-income countries but from a much higher base.

The comparison is not between rich and poor countries today, but between poor countries today and what today's rich countries looked like when they began industrialising. The eight differences are what make "just do what Britain did" unworkable.

This section asks a harder question than the rest of the chapter. Chapter 2 up to now has described how countries differ and measured the difference. §2.6 asks why the differences exist at all — and in particular why some countries are poor for reasons that go back centuries. It is new in the 13th edition and the slides never mention it, but it is the frame the whole field now uses, and it is what Chapter 3's dependence school was fumbling towards without the evidence to back it up.

The section's organising device is Figure 2.9, a schematic with twenty-two numbered arrows connecting climate, colonial history, inequality and institutions to present-day income. The chain underneath it runs like this:

The four families of explanation

The reversal of fortune

This is the section's sharpest empirical point, and it is a falsification test rather than a correlation. If geography were fundamental to development prospects, the argument runs, then the areas that were most prosperous before colonisation should still be the most prosperous now.

The explanation is uncomfortable. Where local populations were larger, denser and more organised, it was easier for colonisers to take over the existing structures and collect tribute — so the institutions built there favoured extracting existing wealth over creating new wealth. Where populations were sparse, taking over was harder and colonists had to settle and produce, which required institutions that encouraged investment: constraints on executives, protection from expropriation.

The mechanism for how this persisted is the difference between formal and informal rules. As North stresses, a constitution can be rewritten in a day; the informal norms around it — how people expect contracts to be honoured, whether officials are expected to take bribes — "usually change only ever so gradually". The Congo is the textbook's illustration: the rule of Leopold II is arguably an ultimate cause of the Mobutu regime after independence.

Engerman and Sokoloff: endowments, inequality, and the two Americas

The second strand explains why the New World diverged internally. What the land could grow determined who worked it:

The mechanism connecting inequality to slow development has a name: the social conflict theory of institutions. Where inequality was extreme there was less investment in human capital and in public goods, and less movement toward democratic institutions — with causality running both ways, since education and institutions reinforce each other. The textbook is careful to add that this is not a story about the English being gentler colonisers, given the treatment of Native Americans and of slaves in the southern colonies.

Section B is the calculation section — the two HDI worked examples and the PPP conversion. Section C reproduces the instructor's own multiple-choice questions.

Sustained growth is recent

Sustained increases in standards of living are a recent phenomenon. Modern economic growth emerged only in the most recent two or three centuries. Until about 12,000 years ago humans were hunters and gatherers; the agricultural revolution around 10,000 BC brought settlements and eventually cities, but even the sporadic peaks of achievement that followed were characterised by low average living standards — wages in ancient Greece and Rome were roughly equal to wages in fifteenth-century Britain or seventeenth-century France, all of them well before modern growth began.

The instructor's figure plots average world GDP per capita from 1 to 2010 (Maddison's data). It is flat for most of its length and then turns up sharply. His framing of the timescale is worth remembering: compress the 130,000 years since modern humans appeared into a single day, and the era of modern growth would have begun only in the last three minutes.

The Great Divergence 分化

The magnitudes are the thing to remember:

Since 1700, living standards in the richest countries have risen from roughly $500 per person to approaching $45,000 — a factor of 90 in a period that is a flash in human history.

Modern growth around the world

After the Second World War, growth in Germany and Japan accelerated sharply — Japan averaging nearly 6% per year between 1950 and 1990 — and both settled at roughly three-quarters of the US level. The United Kingdom was the richest country in the late nineteenth century, then slipped because it grew substantially slower than the United States; since 1950 the two have grown at more or less the same rate, with UK income staying at about three-quarters of the US level. Brazil accelerated until 1980 and then stagnated; China and India have had the reverse pattern.

In per capita terms the ratios are: Japan and the United Kingdom about 3/4 of the United States, Brazil and China about 1/5, and Ethiopia only about 1/40.

Reasons to expect convergence

1. Technology transfer Enables developing countries to leapfrog earlier stages — no need to rediscover the steam engine
2. Diminishing returns to factor accumulation An extra machine adds more output where machines are scarce — so the poor should grow faster

The slide's own caution: divergence occurred for two centuries from the start of the industrial revolution, but the most recent data demonstrate that, on average, (re-)convergence is now underway. "Expect convergence if conditions are similar" is the qualification that later becomes conditional convergence.

Four periods of the convergence record

Growth convergence ≠ absolute income convergence

This is the chapter's sharpest empirical point, and it is easy to get wrong. Over 1990–2017:

China and India grew far faster than the high-income OECD countries. But because they started from such a low base, the absolute gap in per capita income between them and the advanced economies still widened substantially. Relative convergence — closing the ratio — is not the same thing as absolute convergence, which would mean closing the difference.

Data source on the slide: Penn World Table.

A broad sample of countries

Over 1960–2007, growth rates across countries ranged from −23% to +16% per year. Some countries exhibited a negative growth rate; others sustained nearly 6%; most sustained about 2%. Per capita GDP in 2007 varied by a factor of about 64. The lesson the slide draws is the familiar one: small differences in growth rates result in large differences in standards of living over a few decades.

The chapter organises the field into four families. Each is a claim about what causes underdevelopment, and therefore about what should be done.

Rostow's stages of growth

The classic statement of the linear-stages view: the transition from underdevelopment to development can be described in terms of a series of steps or stages through which all countries must proceed. The sequence is what gives the family its name, and the assumption that the sequence is universal is what the later critiques attack.

The Harrod-Domar growth model

The model the stages view needed to make its arithmetic work. Its appeal to scholars and politicians is that it is very simple — and its origin is the Marshall Plan of 1948–51, where the question was precisely how much investment a country needed to hit a growth target. Two economists developed it independently in the 1940s: Roy Harrod (England) and Evsey Domar (MIT).

The derivation

Start from the identity between saving and investment. If the saving rate is s and the capital-output ratio is c — the amount of capital needed for one unit of increase in GDP — then:

Saving equals the saving rate times income; that saving finances investment; investment adds to the capital stock; and the capital required is c times the increase in output.

Rearranging (3.6) gives the model's central result — the rate of growth of GDP:

In the absence of government, the growth rate of national income will be directly related to the savings ratio and inversely related to the economy's capital-output ratio. That is the whole policy content of the model: to grow faster, save more or make capital more productive.

What determines c?

The slide pushes on the capital-output ratio, since it is the more interesting of the two parameters.

Empirically, over 1970–2019 capital-output ratios followed clearly divergent patterns across economies (Penn World Table): Japan and South Korea saw a steady long-term increase driven by sustained capital deepening; the US ratio remained relatively stable; most Latin American and African economies spiked sharply in the 1980s amid economic distress and then declined; India and Indonesia dipped in the 1980s before returning to a gradual upward trend.

The fixed-coefficient production function

The model's assumptions are easiest to see in the production function. Capital (K) and labour (L) are always used in a fixed proportion; there are constant returns to scale, so doubling both doubles output. The slide's example: to produce 100 tons of cement a year a country needs $10 million of capital and 100 workers.

With that functional form, the isoquants are right-angled rather than smooth: to produce a given level of output you need at least a certain amount of both inputs, and extra quantities of one alone do nothing. The kink of each isoquant lies on the ray K/L = 1/10, which is why the model implies capital and labour must grow at the same rate.

Criticisms of the stages model

The instructor's note adds the practical objection: in reality, more saving and investment is not a necessary condition for economic growth.

The structural-change family studies the transformation of an agricultural economy into a more industrialised one, using the tools of neoclassical price and resource-allocation theory and econometrics. Two landmarks: W. Arthur Lewis's "two-sector surplus labour" theory, and Hollis B. Chenery and coauthors' "patterns of development" empirical analysis. Lewis — Saint Lucian, Nobel Memorial Prize 1979 — is the one the chapter develops in detail.

The agricultural sector

Fixed capital K_A and unchanging technology t_A, so production varies only with labour input L_A. Four properties:

1. Decreasing returns in terms of labour Adding workers to fixed land adds less and less output
2. Zero return to labour after a certain point This is surplus labour: production does not increase with extra labour
3. All workers share the production equally So the wage equals the average product, not the marginal product

Mechanically, the story runs: draw a production function that becomes flat at a certain point; from it derive the MP curve, which is also the labour demand curve and which turns zero at the same point where the TP curve becomes flat. There is no labour supply curve — employment is up to the point where MP becomes zero, and the wage is determined by the average product, because there is no labour market in the conventional sense. It is a subsistence economy, and W_A = TP_A / L_A.

The industrial (modern) sector

The same construction, with two changes. Draw a production function; derive the MP curve, which is again the labour demand curve. But the labour supply is horizontal — perfectly elastic — and the wage in the modern sector is higher than the agricultural wage. Labour supply and labour demand together determine the wage level and employment in the industrial sector.

The reason the supply is horizontal is the whole point: while surplus labour exists in agriculture, the modern sector can hire as many workers as it wants at the going wage, because those workers are producing nothing where they are. The wage only has to be slightly above the agricultural wage to attract them.

The two sectors together: migration and industrialisation

Put the panels side by side and the model becomes a theory of development. Modern-sector growth raises the demand for labour; the higher wage pulls workers out of agriculture; the modern sector expands and the economy grows. This self-sustaining growth will not last forever, though:

1. Surplus labour is eventually absorbed Once the modern sector has hired everyone who was producing nothing
2. Further withdrawal reduces agricultural output Because the next workers to leave were producing something
3. The marginal product of agricultural labour rises above zero More land per person remaining
4. Cheap labour is gone; the industrial wage must rise Which means an upward-sloping labour supply curve

Criticisms of the Lewis model

The dependence school denies the premise of both families above: that development is a matter of getting the quantities right. It argues that developing countries are held where they are by external and internal institutional and political constraints — and that the obstacles are therefore structural, not technical.

The neocolonial dependence model 新殖民主义

An indirect outgrowth of Marxist thinking. Its elements:

The dualistic-development thesis 二元发展论

The second strand, which is about structure rather than power. Four propositions:

1. Superior and inferior elements can coexist A modern sector and a traditional one can exist side by side in the same economy
2. The coexistence is chronic It does not resolve itself over time; it is a stable state, not a transition
3. The degrees of superiority or inferiority tend to increase The gap widens rather than narrowing
4. The superior element does little or nothing to pull up the inferior Which is the direct contradiction of the Lewis model's prediction

Criticisms and limitations

The instructor's note records the school's own prescription in one word: autarky is best — cutting the ties to the core rather than negotiating better terms within them.

The counterrevolution challenges the statist 计划经济 model — the assumption, shared by the stages models and the dependence school alike, that the state is the agent of development. It comes in three flavours, which the slide distinguishes carefully.

Main arguments

1. Denies the efficiency of intervention Planning does not improve on the market outcome
2. Points up state-owned enterprise failures Public firms are inefficient — the case for privatisation
3. Stresses government failures Rent-seeking, capture, and the political economy of policy

And one theoretical point, which is the bridge to the Solow model: traditional neoclassical growth theory — with diminishing returns — cannot sustain growth by capital accumulation alone.

Developed in the mid-1950s by Robert Solow of MIT, and the basis for the Nobel Prize he received in 1987. The slide's framing of the question it answers is worth keeping: in 1960 South Korea and the Philippines were similar in many respects — per capita GDP about $1,800 and $2,200 respectively, less than 15% of the US level; populations of about 25 million, half of working age; similar fractions working in industry and agriculture. Between 1960 and 2007 their paths diverged dramatically: the Philippines grew at about 1.7% per year, South Korea at just under 6%, reaching nearly $24,000 against the Philippines' $5,000. The Solow model is the starting point for understanding that difference.

What it adds to the production model

Assumptions: the neoclassical production function

The Cobb-Douglas production function

Most popular function form for production. Assume A is determined exogenously; it has decreasing returns to capital.

Note that the slides use the square-root case specifically: both exponents are 0.5, so α = 0.5 throughout. That matters for the steady state below.

Dividing through by L: the per-worker form

The model is solved in per-worker terms. Divide both sides by L:

And in general form: y = Af(k).

The capital accumulation equation

This is the second equation of the Solow model:

s = saving rate · y = output per worker · n = population growth rate · d = depreciation rate

The growth of the capital-labour ratio relies on three things: savings, population growth and depreciation. The two equations — y = Af(k) and Δk = sy − (n+d)k — together determine the equilibrium.

The Solow diagram

Three lines on axes of k (horizontal) and y (vertical): the production function y = Af(k), the saving curve sy below it, and the straight line (n+d)k from the origin.

At the steady state

k and y are constant. Are K and Y also constant? No. Remember that Y = yL and K = kL, and L is growing at a rate of n. Therefore Y and K grow at a rate of n. The instructor's example: if y = 10 and L is growing at 2%, then Y is growing at 2%.

The closed form is worth knowing, because it makes the comparative statics immediate. Setting sy = (n+d)k and substituting y = Ak^α:

Different initial conditions: the catch-up argument

Suppose two countries X and Z have the same A, f, s, n and d, but country X has larger k and y than country Z. Which country grows faster?

Can a country grow in the long run?

Short-run versus long-run is the whole content of this question. In the short run the economy may not be at the steady state, and k and y may grow. In the long run the economy reaches the steady state, and k and y are stable. So capital accumulation cannot sustain growth — and the chapter is explicit that this is a drawback, because empirically economies appear to continue growing over time.

Strengths and weaknesses of the Solow model

Countries will be rich to the extent that they have a high rate of investment, a high TFP level, and a low rate of depreciation. The model's own verdict on capital accumulation is blunt: because of diminishing returns, an economy accumulating capital sees the marginal product of capital decline, until eventually the additional output produced by investment is only just enough to offset wear and tear. At that point growth stops.

The test case is Robert Barro, 1991, "Economic growth in a cross section of countries," Quarterly Journal of Economics — a paper using regression techniques, on 98 countries over 1960–1985, asking whether incomes converge.

Why no unconditional convergence?

1. The Solow model assumes the same steady state for every country If countries differ in s, n, d or A, they have different steady states and there is no reason for them to converge to each other
2. Many other variables affect growth Such as initial human capital — which is why Barro's second regression controls for them
3. A causality problem in the statistics Is it that saving causes growth of GDP, or growth of GDP causes saving?

The instructor's closing slide on this section shows the growth rate of labour productivity in constant 2017 USD for China, Germany, Japan and the USA — the same comparison at the productivity level rather than the income level, and with the same pattern of faster growth in the catch-up economies.

The chapter closes by refusing to pick a winner. The conclusions the slides draw:

Relevance of growth theories to today's developing countries

The slide is candid about the limits of everything in the chapter:

Two exercises were worked in class and appear at the end of the deck with full answers. They are the closest thing available to a worked exam question, and both are reproduced here verbatim.

The calculation sections reproduce the instructor's two exercises with his own answers. Work them before revealing.

The chapter opens by naming a school of thought rather than a model — one that treats the obstacles to development as failures of coordination rather than failures of resources.

Two things about the definition are easy to miss, and both are examinable.

• The benchmark is another equilibrium, not an optimum. A coordination failure is not simply a bad outcome. It is a bad equilibrium — a situation nobody has an individual incentive to leave, even though everyone would be better off if they all left together.
• Complementarity is what makes it possible. In a competitive market with excess demand, price rises and restores equilibrium. Complementarity defeats that counter-pressure: the more agents who act, the more attractive acting becomes, so the system can settle at either end of a range instead of at a unique point.

The diagram is one of the two pictures this chapter asks you to be able to draw. Read it as follows.

The axes

The horizontal axis is the expected decision by other agents — for concreteness, the average investment level they are expected to choose. The vertical axis is the privately rational decision of a single agent: the level of investment that agent finds it best to choose, given what it expects everyone else to do. The 45° line is the set of points where actual individual investment equals expected investment by other agents — expectations are fulfilled, so what was expected is exactly what turns out to be privately rational.

Why the curve is S-shaped

The function is called the privately rational decision function, and its shape is the theory. It first rises at an increasing rate, then at a decreasing rate. The update states the reasoning as a slow start, a snowball, and diminishing gains — and the reasoning is the examinable part:

1. Some agents invest alone. Even if nobody else invests, a small number of agents will — which is why the curve has a positive vertical intercept (marked Y₁ in Figure 4.1).
2. Early spillovers are weak. With only a few participants, each early investor is relatively isolated from the others, so spillovers are weak and the curve rises slowly.
3. Then a snowball effect. Once enough agents invest, strong spillovers speed up the curve — many investors begin to provide benefits to their neighbours, and the curve increases faster.
4. Then diminishing gains. After most potential investors have benefited, the most important gains have been realised and the growth of the curve slows again.

Equilibrium, and why two of the three are stable

An equilibrium is a point where the privately rational decision function crosses the 45° line — where the investment level expected equals the level all agents find best. Figure 4.1 has three such points, and they are not equivalent. The slides give the criterion purely geometrically, and it is worth memorising in that form:

The textbook adds the expectation-adjustment story that gets the economy to these points in the first place. Suppose firms initially expect no other firm to invest, but some do anyway. Seeing that, firms revise their expectations upward, matching them to the investment they actually observe — and at that higher expectation they want to invest even more. The process continues until the level of actual investment just equals the level expected. That is the equilibrium condition.

What the model is really claiming

The economic content is in the last four slides of the section, and it is stronger than the diagram suggests. Economic development requires coordination: one investment depends on other investments. All are better off with more investors or higher rates of investment — but the market may not get there without proper government policy. Jointly profitable projects may not happen without coordination, which leads to multiple equilibria: either good or bad outcomes. Many least developed countries are caught in such low-investment circumstances, and the slide names sub-Saharan Africa as the example.

The Big Push is the coordination-failure argument made concrete enough to draw. It was pioneered by Paul Rosenstein-Rodan and addresses the obstacles to launching industrialisation within a subsistence economy. The update's framing is that 市场失灵 — market failure — creates a need for public policy intervention, and the specific market failure at issue is a pecuniary externality (货币外部性): a spillover effect on an agent's costs or revenues. The slide's example: an influx of urban residents buying second homes in rural areas bids up local housing prices, making home purchases unaffordable for local young people.

The update draws the link to the model explicitly: one firm's industrial investment expands aggregate market demand, raising the potential profits of other firms. That is what creates investment complementarities — and therefore coordination failure and multiple equilibria. The externality runs through demand, not through technology: no firm's production function changes when another firm invests, but the market its output can be sold into does.

The six assumptions

The instructor lists the model's assumptions explicitly, and they are a plausible exam question in themselves.

Reading Figure 4.2

The update adds a two-line reading guide. Point A is the outcome when only one modern firm enters and the rest of the economy stays traditional. Point B is the outcome when all N sectors modernise together, so that demand has expanded. Everything else in the figure is a comparison of those two scales.

Two production functions, and a family of wage bill lines:

Price is 1, so revenue PQ can be read straight off the Q axis — which is why nothing in the figure needs a money axis. The horizontal axis is labour, and two points on it matter. L/N is the labour available to each of the N product sectors. F is the modern firm's fixed labour requirement, so the modern production function simply starts at F instead of at the origin — that horizontal segment from 0 to F along the axis is the increasing-returns assumption, drawn.

The update then states the conditions for multiple equilibria as a question about point A: whether the modern firm enters there depends on two things — (1) how much more efficient the modern sector is than the traditional sector, and (2) how much higher modern wages are than traditional wages. The two pull in opposite directions. Greater efficiency (a steeper technique, a smaller c, or a lower fixed cost F) pushes the wage-bill line down, below A; a higher modern wage W pushes it up, above A. The model's whole outcome turns on which force wins.

The three cases

Whether industrialisation happens depends entirely on where the wage bill line falls. The slides work through three positions; the labels W₁, W₂, W₃ name three different economies, not three moments in time.

Key results: what keeps B viable

The update closes the section with four conditions. They are worth memorising as a set, because each is a way the good equilibrium could fail to exist.

• Limit pricing. Modern firms cannot raise price above P = 1, or traditional firms re-enter and undercut them. So B stays profitable only because modern costs are low enough at P = 1.
• The closed economy is essential to the mechanism. With free exports, one firm could sell abroad without the other sectors modernising, so the demand externality disappears — and with it the multiple equilibria.
• Workers gain. They earn at least as much as in the traditional sector, can buy more goods, and move voluntarily — so industrialisation is a Pareto improvement, not a redistribution at someone's expense.
• Demand is sufficient. All national income (wages + profits = output per good × N) is spent on domestic output, so the expanded supply at B finds a buyer.

To summarise, in the slides' own words

• Whenever the wage bill line passes below point A, the market will lead the economy to modernise; whenever it passes above A, it will not.
• The steeper (more efficient) the modern-sector production technique, or the lower the fixed costs, the more likely the wage bill passes below the corresponding point A. In the model's parameters: a smaller c or a smaller F.
• If the line passes above B, it makes no sense to industrialise.
• With a wage like W₂ there are two equilibria — one industrialised, with higher profits, wages and output; one not, with lower wages and output. The higher-output equilibrium is better, but the market stays trapped in the bad one because of a coordination failure — no firm wants to move first. So synchronised investment (a Big Push) and a role for policy are needed to start development.

Four further effects

The wage mechanism is one reason a big push may be needed. The slides list four more, and they are distinct enough to be listed separately in an answer.

Why a super-entrepreneur cannot solve it

The obvious objection is: if all these investments are jointly profitable, why doesn't one agent capture the rent by doing them all? The slides' title for this slide is blunt — why the problem cannot be solved by a super-entrepreneur — and the list of reasons is the answer.

• Capital market failures — the necessary funds cannot be raised at any reasonable price.
• The cost of monitoring managers, which is an asymmetric information 信息不对称 problem: one party to a potential transaction has more information than another.
• Communication failures — the coordination problem reappears inside the organisation.
• Limits to knowledge — no single agent knows what all N sectors require.
• And no empirical evidence that it would be possible. The slides close the list with an appeal to the record rather than to theory.

In a nutshell: the mechanisms

The chapter's own summary of what industrialising firms do for each other. Firms adopting increasing-returns-to-scale technologies have the following effects:

• Raising total demand — which is the Case 3 mechanism.
• Reducing fixed costs of later entrants — infrastructure, standards, supplier networks.
• Redistributing demand to later periods when other industrialising firms sell. This is the intertemporal effect.
• Shifting demand toward manufacturing goods, usually produced in urban areas. This is the urbanisation effect.
• Helping defray 分摊 the costs of essential infrastructure — a mechanism that holds equally for training costs and other shared intermediate inputs. This is the infrastructure and training effect.

The slide's closing sentence states the common structure: each of these has external beneficial effects on other industrialising firms. Economies of scale mean that the first mover cannot capture all the returns from its own investment — which is the definition of a positive externality, and the reason the market under-provides it.

Four shorter arguments, each identifying another route to a bad equilibrium. They are listed rather than modelled, but they are all examinable as named items.

Note also that linkages is the one item in this section that is a policy proposal rather than a diagnosis, and it is the bridge to the last two sections: pick the industries with the right linkages, and you may be able to trigger the good equilibrium with a much smaller intervention than a full-scale big push.

Everything to this point has run through demand: your return rises because other firms' investments create a market. Kremer's model reaches a trap through a completely different channel — the technology of production itself — and it is the second picture this chapter is likely to test.

The name

The model takes its name from the 1986 Challenger disaster, in which the failure of one small, inexpensive part — an O-ring seal — caused the space shuttle to explode. The name is the argument: when tasks are complements, the lowest-quality input governs the value of the whole. Kremer's model explains not only the existence of poverty traps but also why countries caught in such traps may have such exceptionally low incomes compared with high-income countries — the shortfall is multiplicative, not additive.

Setting up

Production is modelled with strong complementarities 互补性 among inputs. Suppose a production process is broken into n tasks, ordered strictly by the level of skill q required, with 0 ≤ q ≤ 1. The higher the skill, the higher the probability that the task will be "successfully completed". The assumptions are that firms are risk-neutral, labour markets are competitive, and workers supply labour inelastically.

Output is given by multiplying the q values of the n tasks together, in turn multiplied by a term B that depends on the characteristics of the firm and is generally larger with a larger number of tasks. If each firm hires only two workers, the O-ring production function is:

Positive assortative matching

The most prominent feature of this production function is positive assortative matching 正向匹配: workers with high skills work together and workers with low skills work together. You will be more productive when working with a more productive person. The slide then sets up a four-person economy with an inequality:

Read this as a lemma, not as a comparison of two arrangements. It is the statement that (q_H − q_L)² > 0, which is true whenever the two skill levels differ at all — the textbook says so explicitly ("recall that (x − y)² > 0 for any x that is not the same as y"). It rules out the badly matched pairing but does not by itself rank the two good arrangements against each other. The actual claim being established is that total value is higher under skill matching than under skill mixing:

The consequences follow. Because total value is higher when skill matching rather than skill mixing takes place, the firm that starts with high-productivity workers can afford to bid more to get additional high-productivity workers, and it is profitable to do so. A cascading process follows: the highest-skill workers pair off first, then the next highest, and so on, until the least productive are left with each other. The textbook's illustrations are worth remembering because they make the mechanism intuitive — a symphony orchestra is adversely affected as a whole by hiring one single poor performer, so an excellent orchestra has every incentive to bid the most for an outstanding performer; the restaurant with the very best chef hires mature, highly trained, full-time waiters, while a fast-food restaurant does not hire a famous chef. The textbook also offers the analogy to Gary Becker's "marriage market" model.

The numerical example

This is the calculation most likely to appear on a problem set, and it must be reproduced exactly.

Implications for development and the distribution of income

The slides list four implications, and the textbook adds a fifth:

• Firms tend to employ workers with similar skills for their various tasks.
• Workers performing the same task earn higher wages in a high-skill firm than in a low-skill firm.
• Because wages increase in q at an increasing rate, wages will be more than proportionally higher in developed countries than would be predicted from standard measures of skill. This is the model's answer to why measured human capital explains so little of the income gap.
• When those around you have higher average skills, you have a greater incentive to acquire more skills — the individual's human-capital decision depends on other workers' decisions.
• One can get caught in economy-wide low-production-quality traps. There could thus be a case for an industrial policy to encourage quality upgrading, and the slide flags this as relevant to a country trying to escape the middle-income trap 中等收入陷阱.

Finally, two consequences about bottlenecks 瓶颈, which are the O-ring logic applied to the weakest link:

• O-ring effects magnify the impact of local production bottlenecks, because such bottlenecks have a multiplicative 倍数的 effect on other production.
• Bottlenecks also reduce the incentive for workers to invest in skills, by lowering the expected return to those skills — the trap closing itself.

Hausmann and Rodrik's objection is aimed at the advice developing countries are usually given, and it is an objection about information.

It is not enough to tell a developing country to produce "labour-intensive products", because there are thousands of them. The advice does not say which. That is the problem: a country's true costs of producing any particular good — direct and indirect — are not knowable in advance. They are discovered only by someone trying, and once someone has tried, the answer is public. The discoverer bears the cost of the experiment; the imitators get the answer for free. That is an information externality, and it means too little discovery happens.

The policy conclusion is that industrial policy may help to identify the true costs of potential products to specialise in, by working in two stages:

• First stage: encourage exploration.
• Second stage: encourage movement out of inefficient sectors and into more efficient ones.

Note the shape of that: the state's job is not to pick winners and keep them, but to sponsor the experiment and then to accept the answer.

The three building blocks

The slides give three building blocks of the theory and a case example for each, and the examples are the part to memorise — they are the evidence that the theory describes something real.

The chapter closes its constructive part by refusing to give a general answer — and the refusal is the framework.

• Focus on a country's most binding constraints 绑定约束 on economic growth. A binding constraint is the one limiting factor that, if relaxed, would be the item that accelerates growth.
• There is no "one size fits all" in development policy.
• It therefore requires careful research to determine the most likely binding constraint.
• The framework has been adopted by the Inter-American Development Bank (IDB) and the World Bank — which is why it is worth knowing by name.

Figure 4.3 — the decision tree

The tree starts from a single symptom and works backwards. The problem at the top is low levels of private investment and entrepreneurship. There are only two possible causes, and they are an exhaustive pair:

The bottom row of the tree names the possible binding constraints, and it is the part worth learning as a list:

4.8 Conclusions

The deeper point is that people keep doing inefficient things because it is rational to keep doing them. This leads to a fundamental problem of coordination failure. Government policy and aid will be necessary to overcome the resulting vicious circles of underdevelopment.

The chapter's final sentence is a warning as much as a conclusion. Government is not an outside party that arrives to fix the equilibrium: it is part of the economy being modelled, and its behaviour is itself partly determined by the underdeveloped economy. The textbook's example is a dictator who prefers to keep his country in an underdevelopment trap, knowing that as the economy develops he will lose power. The same complementarities that make policy powerful make a bad policy durable — and that is why the chapter ends on analysis rather than on prescription.

Two exercises appear on the closing slide, and the reading he set alongside them. Both are drawing questions, which is a reliable signal about what the exam will ask.

This chapter is unusually well supplied with margin definitions — the textbook defines 25 terms inside its own pages. Every term below with a 中文 anchor is one the instructor glossed himself on the slides.

The O-ring calculation questions reproduce the instructor's own numbers exactly. Work them before revealing.

The chapter opens with seven questions, and they are worth reading as a list because the rest of the deck answers them in order: what is the extent of relative inequality and how is it related to poverty; who are the poor; who benefits from growth; does rapid growth necessarily cause greater inequality; do the poor benefit from growth; are high levels of inequality always bad; and what policies can reduce poverty.

Two ways to describe a distribution

An individual's position on the size distribution is called a quintile or decile. The first decile is the poorest tenth, the second decile the next tenth, and so on. The instructor's own gloss is worth remembering: quintile 五分之一, decile 十分之一.

Table 5.1 — the raw data the chapter works from

The table arranges twenty individuals in ascending order of annual income, from 0.8 units to 15.0 units, and reports the share of national income held by each fifth and each tenth. Read the share columns, not the individual incomes: the question is always what fraction of the total a group holds.

Two summary statistics come straight out of the table.

• The Kuznets ratio 库兹涅茨比率 is the income of the top 20% divided by the income of the bottom 40%. In Table 5.1 the top quintile holds 51% and the bottom two quintiles hold 5 + 9 = 14%, so the ratio is 51 / 14 = 3.64. A larger ratio means a more unequal distribution.
• The shares must add to 100. The quintile column gives 5 + 9 + 13 + 22 + 51 = 100 and the decile column gives 1.8 + 3.2 + 3.9 + 5.1 + 5.8 + 7.2 + 9.0 + 13.0 + 22.5 + 28.5 = 100. This is a useful check on any transcription of the table; a column that does not sum to 100 has been read wrong.

The individual incomes in the underlying table are not needed for the exam, but one feature of them is worth noticing: the jump between individual 16 (7.1) and individual 17 (10.5) is larger than the entire range from individual 1 to individual 10. The top of a developing country's income distribution is not simply a scaled-up version of the bottom; it is a different distribution.

The Lorenz curve is the picture behind every inequality number in the chapter, and it is drawn from exactly the cumulative column of Table 5.1. The horizontal axis is the cumulative percentage of income recipients, from poorest to richest; the vertical axis is the cumulative percentage of income they receive.

Reading the curve

• The 45° line is perfect equality. If everyone had the same income, the poorest 10% would receive 10% of income, the poorest 20% would receive 20%, and so on. The curve would be the diagonal.
• The Lorenz curve always lies on or below the diagonal, because the poorest group cannot receive more than its proportional share in a distribution ranked from poorest to richest. It starts at (0, 0) and ends at (100, 100).
• The further it sags, the more unequal the distribution. Figure 5.2 puts a relatively flat curve beside a deeply bowed one and asks you to see that the second is the more unequal economy — without any numbers at all.
• A curve that crosses another is ambiguous. Figure 5.4's curves B and C cross. B lies above C for the poorest deciles but below C in the middle, so which economy is "more equal" depends on which part of the distribution you care about. The slides take the sensible line: if you prioritise the poorest, B is more equal because the poorest are richer; if you think a strong middle class is what equality means, C is. A and D do not cross, so A is unambiguously more equal than D.

The Gini coefficient

The Gini coefficient 基尼系数 is the most widely used single number. Geometrically it is the area between the Lorenz curve and the 45° line (call it A), divided by the area of the whole triangle below the 45° line (A plus B):

It runs from 0, perfect equality, to 1, perfect inequality. A larger coefficient means more inequality. For Table 5.1 the slide computes the area under the Lorenz curve by adding up trapezoids between successive quintiles:

China's Gini coefficient

The deck tracks the Chinese series, and the shape is the interesting part: it rose steeply during the reform era, peaked around 2008, and has edged down since.

The deck gives three reasons for the large income gap, and they map onto the Lewis model from Chapter 3: migration of rural labour into a dual economy changes the distribution as it proceeds; uneven development across industries widens the gap between them; and unbalanced regional development widens the gap between regions. All three are transitional mechanisms — they say inequality rises because the economy is moving, not because it is stuck.

The size distribution asks who gets income. The functional distribution asks which factor gets it: labour in wages, capital in profits. The neoclassical picture is a demand-and-supply diagram with two factors, capital fixed and labour variable.

• The labour demand curve is the marginal product of labour, MP_L, sloping down because of diminishing returns to labour on a fixed capital stock.
• The labour supply curve S_L slopes up; where the two meet gives the employment level L_E and the wage W_E.
• Total output is the whole area under the demand curve up to L_E, because each additional worker produces their marginal product and the area sums those products.
• Labour's share is the wage rectangle 0–W_E–E–L_E: the wage times the number of workers.
• Capitalists' share is the triangle above the wage and below the demand curve, W_E–R–E: output minus the wage bill.

The slide ends with its own limitation, and it is the one to reproduce: this theory ignores nonmarket forces — institutions, bargaining power, discrimination, and policy. That is not a small caveat. In a developing country, a large informal sector, weak labour law, caste or ethnic stratification, and government-set wages all break the neat tangency. The functional distribution is a benchmark, not a description.

Inequality is about the shape of the whole distribution. Poverty is about the bottom of it, and it needs its own measures because two countries with identical Lorenz curves can have very different numbers of people below a poverty line. The slides build the measures in a sequence, and each one repairs a defect in the one before.

The headcount index

The headcount index 人头指数 is the fraction of the population that is poor. Its defect is that it is completely insensitive to depth: a person earning $0.50 a day and a person earning $1.85 a day are counted identically under a $1.90 line, although the first is far worse off. It also creates a perverse incentive for officials, who can show progress most cheaply by lifting the people nearest the line.

The total poverty gap

The total poverty gap (TPG) sums, over every poor person, the shortfall between the poverty line Y_p and that person's income Y_i. It is the amount of income that would have to be transferred to bring everyone up to the line, ignoring administrative costs and behavioural responses. Figure 5.6 draws two countries with the same 50% headcount and different gaps: the one whose poor are further below the line has the larger shaded area, and needs the larger effort.

Four ways to normalise the gap

The FGT index

The Foster–Greer–Thorbecke index puts the whole family in one formula:

• α = 0 gives the headcount ratio H/N.
• α = 1 gives the normalised per-capita poverty gap, P₁ = (H/N) × NIS.
• α = 2 gives the squared poverty gap P₂, which weights the deepest shortfalls most heavily. This is the measure the World Bank and Mexico's Progresa programme use.

The slide gives the alternative form of P₂, which is worth knowing because it shows what the measure is made of:

where CV_p is the coefficient of variation of income among the poor. So P₂ rises with the headcount, with the depth of the gap, and with inequality among the poor.

The Multidimensional Poverty Index

Income is a means, not an end. The MPI follows Sen's capability argument from Chapter 1 and counts deprivations directly. Identification uses a dual cutoff: first, a cutoff within each dimension (analogous to a poverty line); second, a cutoff in the number of dimensions in which a person must be deprived to count as poor. The deck's weights are:

With H the multidimensional headcount and A the average intensity of deprivation, the adjusted headcount ratio is simply their product:

The deck's Table 5.2 makes the multiplication visible. Bangladesh 2019: H = 0.383, A = 0.394, so MPI = 0.383 × 0.394 = 0.151. Afghanistan 2022/23: 0.522 × 0.514 = 0.268. Niger 2012: 0.899 × 0.661 = 0.594. The product form matters because it separates how many people are poor from how intensely they are deprived.

Why extreme inequality is bad

The slides give three reasons, and they are of different kinds. Extreme inequality leads to economic inefficiency — it blocks the poor from credit, education and health investment, so productive capacity is wasted. It undermines social stability and solidarity, which makes policy harder to sustain. And it is unfair — a normative judgement the chapter does not pretend is derivable from efficiency alone.

The welfare function the chapter uses stacks the three concerns together:

Income per capita enters positively; inequality and absolute poverty enter negatively. The function is a way of keeping the three ideas separate before asking how a policy moves them. It is not a claim that one dollar of redistribution is worth one dollar to the rich: the negative sign on I is exactly the statement that a mean-preserving spread lowers welfare.

Three growth typologies and the shifting Lorenz curve

This is the most examinable diagram in the second half of the chapter. A two-sector economy can grow in three stylised ways, and each has a different effect on the Lorenz curve.

The crossing in the third case is the point, and the slide spells out why. The poor who remain in the traditional sector have the same incomes, but those incomes are now a smaller fraction of a larger total, so the new curve lies below the old one at the lower end. Each modern-sector worker earns the same absolute income as before, but the richest group's share has fallen, so the new curve lies above the old one at the upper end. Two curves that cross cannot be ranked by the Gini coefficient alone, and the slide's answer — "ambiguous" — is the right one.

The Kuznets inverted-U

Simon Kuznets hypothesised that inequality rises in the early stages of growth and falls in later stages, tracing an inverted U against income per capita. The mechanism is structural change: early growth concentrates in a modern sector that employs few people, so the gap between sectors widens; later, the modern sector absorbs enough of the population that the gap narrows again.

The slides immediately qualify the hypothesis with their own data:

• Table 5.4 shows that per-capita income and the Gini coefficient are not tightly related. Malawi (income 320, Gini 44.7) and Niger (360, 34.3) differ sharply at similar incomes; Norway (76,160, 27.5) and the United States (59,160, 41.5) differ sharply at the top. The slide's own caption: "PC incomes are not necessarily related to inequality."
• Figure 5.11 fits a Kuznets curve to a scatter of countries and highlights the Latin American observations, which sit far above the fitted line — high inequality at every income level.
• Figure 5.12 plots the time path for selected countries. Brazil (labelled 1) traces the inverted-U; Hong Kong (4) and Singapore (5) trace a U — inequality fell and then rose. A hypothesis that predicts one shape and is contradicted by countries tracing the other is a tendency, not a law.

Growth and the incomes of the poor

Figure 5.10 plots the growth rate of GNP against the growth rate of incomes of the bottom 40% for a set of developing countries. Points on the 45° line mean the poorest 40% grew at the same rate as the economy; points above it mean they grew faster. Taiwan and South Korea sit above the line, India and the Philippines below. The lesson is the one the instructor keeps returning to: the poor can benefit from growth, but whether they do is an empirical question about the pattern of growth, not an implication of growth itself.

The long-run story is one of extraordinary progress, and the short-run story is one of a setback.

• 1990: 38% of the world's population lived below the international extreme-poverty line.
• 2022: about 9%, or roughly 713 million people, below $2.15 a day at 2017 PPPs.
• 2020: COVID-19 caused the first rise in decades — 8.9% in 2019 to 9.7% in 2020, and the World Bank estimates it pushed an additional 97 million people into extreme poverty.

The regional pattern is the one the chapter stresses. East Asia, led by China, saw a dramatic reduction approaching elimination. South Asia saw a large decline but still has substantial numbers. Sub-Saharan Africa made the slowest progress, and most of the world's extreme poor now live there. The deck's Figure 5.14 plots the global headcount ratio from 1990 to 2022 and marks both the COVID spike and the 2022 level; Table 5.5 gives country incidence at two lines.

Read the two lines together. At $1.90, Bangladesh and India look very different (14.8% against 21.2%); at $3.80 — roughly the upper-middle-income line — India is far worse (71.0% against 65.2%). A country's ranking on poverty depends on where the line is drawn, which is a point the chapter makes repeatedly and which the $3.00 revision of 2025 makes urgent.

Poverty is not randomly distributed. The chapter identifies three groups with systematically higher incidence, and the tables make the gaps concrete.

Rural poverty

The rural poverty rate exceeds the urban rate in every country in Table 5.7, often by a factor of two or more. Zambia 2004: 72.0% rural against 53.0% urban. Vietnam 2002: 35.6% against 6.6% — a fivefold difference. The national rate always sits between the two because it is a population-weighted average, which is why a country can look much poorer or much richer than its countryside depending on how urbanised it is.

Women and poverty

The deck lists women as a high-poverty group without giving a table, and the reason is methodological as much as substantive: household surveys typically measure income at the household level, so they cannot see intra-household inequality. Where separate data exist, female-headed households and women in male-headed households are over-represented among the poor, and the chapter connects this to the household theory in Chapter 6 — women's time, education and employment opportunities are the variables that move both poverty and fertility.

Ethnic minorities and indigenous populations

Table 5.8 is the sharpest evidence in the chapter. In every Latin American country listed, indigenous poverty in the early 1990s was far above non-indigenous poverty: Bolivia 64.3% against 48.1%, Guatemala 86.6% against 53.9%, Mexico 80.6% against 17.9%, Peru 79.0% against 49.7%. And the change columns show that over the following decade indigenous poverty barely moved — 0 in Bolivia, 0 in Mexico, 0 in Peru, −15 in Guatemala — while non-indigenous poverty fell by 8, 5 and 25 points in three of the four countries, and rose 3 points in Peru.

The deck divides policy into four areas of intervention and four instruments. The areas say what part of the distribution you are trying to change; the instruments say how.

Two warnings from the chapter are worth carrying into an exam answer. First, a policy can move the functional distribution and the size distribution in opposite directions — a capital subsidy raises output and may raise the top's income while doing nothing for the poor. Second, the instruments are not interchangeable: a progressive income tax does little in a country where most people are outside the tax net, and land reform does little in a country where land is already fragmented. The binding constraint differs by country, which is the growth-diagnostics lesson of Chapter 4 applied to poverty.

China's war on poverty — the deck's worked case

The instructor devotes eight slides to China, and they are best read as a single case study in how a poverty line, a target and a delivery system fit together.