Topics we covered


Lecture 2: Slide 29 is helpful.  Growth is equiproportional, and is also uncorrelated w/ inequality.  Growth is correlated w/ reduced absolute poverty.  Absolute poverty vs. relative poverty.  Surveys give lower data than GDP (people underreport, rich don't answer)---it's conservative: growth estimates say MDG 1 goals are met, whereas surveys say it's not. (Dollar & Kray, Sala-i-Martin)

Lecture 3: Poor live in large families, lots of young.  Have few material possessions, but in some countries many have tvs, radios.  increased mortality rates the poorer you are.  food expenditures are 50-80% for poor, with smaller fraction on alcohol, festivals, durable goods, health, latrine depending on area. % on education increases w/ income.  people have multiple jobs, often in low-income, high-competition fields (no scale).  few migrant workers, usually for short time.  little credit, high rates, no formal insurance. (Banerjee + Duflo)

Lecture 4: Malthus said population growth eventually kills finite resources.  Technology/health proves this wrong.  Diff-in-diff on time before 1700/after 1700 (potatoes in old world) and geography (compare potato suitability for controls).  (Nunn, and Qian).  Poverty trap model (calory capacity curve above/below y=x, equilibria at y=x).  Potatoes shift capacity curve up (output increases for any given income).  Africa hasn't seen potato/green/tomato/corn revolution others have.  Children have iron/iodine deficiencies---long-term poverty trap, famine affects children more than adults.

Lecture 5: Bengal famine: 1.5-7M dead. 3 phases. burma taken, cyclone hits, worries about japanese (scorched earth). even still, 1943 rice output (famine) higher than 1941 (before famine).  inflation led to bubble for rice, suppliers hoarded (rational reaction), and paddy huskers/farm labor got no rice/couldn't afford.  government mismanagement: didn't open trade, helped calcutta w/o helping countryside. poor migrated to city, but no help there.  man-made famine should not exist in democracy. (Amartya Sen)

Lecture 6: Last large-scale famine was Ethipoia '84-'85 (~1M dead).  before that china w/ 15M in 1959-61. phases: immediate deaths, reduced births, long-term effects, especially children/markets. poverty prior to famine->easier to see deaths. war more than food decline in 1900s (Cormac O'Grada---making famine history). China: enough food available even in food availability decline: poor flow of information, quotas pulled grains from farms to cities, killed farmers (Qian + Meng).  India govt. reacts more quickly to disaster when local newspaper is larger (Besley + Burgess). NGOs lag disasters, need money, media coverage helps, competing events (olympics, school shooting) means no coverage, no aid. coverage increases aid, competing stories give less coverage. 

Lecture 7: Giffen goods (Robert Jensen), nutrition-based poverty trap. malnutrition high (1B): MDG 1 says reduce poverty/HUNGER. more money->poor buy other things, or different food. Engel curve (calorie elasticity): 10% increase income==7% spent on food, 3.5% on increase in caloric consumption. giffen good: price decrease, demand decrease. two effects of price: substitution (item gets cheaper than others, you buy more), income (same quantity gives you more to spend elsewhere). substitution stronger when %income on item is small. food for poor sees larger income effect (rice in china). verify that giffen good: high price->high demand, not high demand->increased prices. distribute enough vouchers to reduce price, measure income/substitution effect. hunan: rice is giffen good. guansu: wheat not giffen good. implications->be careful w/ subsidies (india)->may promote bad nutrition. in past 25 years, engel curve has dropped, and child nutrition has not changed. remember: people!=machines, make human choices.

Lecture 8: child labor vs. school. demand: parents need $ now, cost of tuition, transport, uniforms, free meals. supply: transport infrastructure, books, schools, teachers. returns to education are discounted: hard to see benefit today. externality: schooling is also a public good, hard for parents to internalize externality. two poverty traps equillibria (all children at school, all children working) form two reasons (decreased wages w/ more kids, increased fines w/ more child labor).  (Christopher Udry, Kaushik Basu: two child labor papers).

Lecture 9: Cost/benefit of methods school enrollment/consistent attendance. needs assessment: who is targetted, what is environment, what is problem? process evaluation: cost effective, do people follow rules, does target population get service? Impact evaluation: does increase in X change Y? for true impact, must know counterfactual (if you hadn't been there). can't get that, so randomly pick control group. still, impact also includes selection bias, which you must reduce: randomization helps. large enough groups, assume distribution of treatment/control is identical, then impact=tratment effect. process eval. gives costs, impact gives benefits: divide after normalizing for # years of benefit, etc. deworming > school meals > uniforms > remitance, but it's in different countries. (Duflo, JPAL)

Lecture 10: MDG equal boy/girl education, universal education by 2015.  great strides in sub-saharan/southeast asia. education benefits individuals (income, technology understanding, germ theory) augmented by tech., held down by no industry. benefits to society (teach others, adapt to technology) (Anne Case). richer countries more educated, ~10% increase in income/year education. educated countries should be 80% richer, but at 15x---is education really the cause? Easterly suggests reverse causality: rich->education. Easterly: no sign of growth where education recently increased. Response: many ways to measure education (enrollment, consistency, etc.), maybe more education->lower quality, maybe external events (war) stunted growth. so look at individual level, not cross-country. diff-in-diff: indonesia started schools for kids <12 in 1973 in poor areas. measure outcome for educated/uneducated kids in 1995, see .2 years, 1.5% wage increase for each 1000-kid school, or 7.5%/extra year. school building is an IV for education on wages. education: babies less likely to die early, born with low weight. uniforms in kenya: reduce dropout rate, teen pregnancy rate.

Lecture 11: India has low literacy in kids in school, and low # kids in school. can get kids in school, harder to get them to perform well. randomized trial of balsakhi (remedial education) and computer-assisted learning (CAL). balsakhi: cheap, decrease student/teacher ratio, pick up lost students, randomized between grades.  CAL: cheap labor oversees students share computer 2 hours/day to play math games. stratify by balsakhi treatment, gender, math test scores, language of instruction. switched treatment/control after a year. low attrition/no sig. effect on attendance for treatment/control. comparison: subtract last year's mean, divide by stdev. balsakhi: benefits all students, especially those who were taught, by >.2 std devs (often >2), which is big and significant, results seem cumulative, math > language. CAL benefits math not language, but significant as well. CAL is 6.7x as expensive, balsakhi better for language, cal for math, but not by much. effect of unbalsakhi'd students not significant: no class size effect.  (Banerjee, Duflo, ...)

Lecture 12: Study for Seva Mandir on health delivery in rural Udaipur. four parts: village survey (maps, listing infastructure), facility survey, weekly facility visits (are they open?), household survey (income, health, happines, direct health measures). low literacy + 20% have electricity. most underweight, anemic, difficulty walking/lifting/squatting. 7% consumption on health, poor more likely to see private doctor/quack. public facility closer, more trained doctor, and cheaper, but poor go to private doctors/bhopas (not medical) more. often govt. clinics don't have nurses/doctors when they should, and often they pander to what clients want (shot, etc.). correlation between staffed clinics and locals seeking them out. increase staff and monitor more, but nurses broke monitoring machines, got exemptions from supervisors. (Banerjee + Duflo)

Lecture 13: Malaria eradication, effects on children from morbidity (Bleakley).  Micro-study: diff-in-diff within countries for born before/after treatment (and in between), and across regions of high/low malaria (areas within usa, brazil, columbia, mexico). highly affected areas saw more benefit, and longer exposure to eradication saw more treatment. same trends for income, literacy, years of schooling except mexico, which couldn't handle more kids. Now turn to mortality for eradication of fatal diseases (Acemoglu + Johnson).  Macro-study: life-expectancy(LE)->GDP? 1940s transition to eradicate many diseases: measure predicted mortality as IV for life expectancy, w/ pre-intervention/frontier death rate depending on when treatment hit country. LE correlated w/ GDP, but is there causality? First stage: correlate LE and PM.  Reduced form: correlate PM and GDP. LE->GDP effect is RF/First stage correlation. First do RF, first stage for population, find that LE->increase population. Then do RF/first stage on GDP, and find no change in growth (more young work, but more old live longer). So gdp/capita goes down w/ LE.

Lecture 14: Insecticide treated bednets (ITNs) are bottom-up way to eradicate malaria. ITNs have externality: randomized kenyan trial shows helps everyone in 300m. Govt. should subsidize externality. Free? Easterly says no, as paying makes you use it more. So price drop changes demand (more demand), and changes intensity of usage (less usage). demand slope usually downward, but usage intensity can be either (want poor to buy it, or want people to feel sunk cost). Randomized kenyan trial, 20 health centers for expecting mothers, 4 control, 5 free, 5 97.5% subsidized, 3 95%, 3 90%. street price is 80% subsidized. measured sunk cost w/ second-stage randomization (to lower price paid). demand effect huge: Free > cheap >> 90% subsidy. intensity usage effect almost non-existant, caveat: low demand at 90%->too few users. no large sunk cost effect. Free has higher cost/benefit. Jeff Sachs maybe vindicated, but lack of significance of usage intensity makes it hard to tell. randomization->internal validity of causality, but external validity (outside of kenya) is not yet there. (Dupas + Cohen)

Lecture 15: Since 1960s, everywhere but subsaharan africa(SSA) has seen food/person output increase (SSA decrease), and has had some green revolution. Fertilizer has helped them all, not been used in SSA. World Bank says increase infrastrucutre, fertilizer not sustainable, but african countries saw success when subsidized it. india/zambia heavily subsidize fert. (.75-2% of GDP). WB says: increase supply by improving roads, technology, lower tariffs, increase demand by increase credit, education, research, irrigation. (NYT article + World Bank Report). chicago tradition/response from duflo, kremer, robinson: not profitable/yes it is, only works on demonstration farms/randomly tested and works in real farms, farmers don't know benefits/when farmers shown benefits, still don't use it, credit constraints/fertilizer is super-divisible, so why don't they buy small amount? their answer: farmers are present-biased---know it's good, plan to use it next season, but don't follow up (it's good for me in future, but not _now_). ran several SAFI experiments, offering delivery now vs. harvest time, subsidize vs deliver, etc., and find present-bias.

Lecture 16: Why don't banks lend to poor?  Overhead/information assymetry. Two reasons for finance: consumption smoothing/risk diversification (no growth), and investment from money-havers to idea-havers (growth. poor mostly want consumption smoothing, but microcredit also allows growth. Even w/o infrastrucutre, poor find ways: diversify income (multiple jobs), save (stock up food, money in matress). Also work together: non-synchronous income shocks, different consumption timing (dowry, etc.), different job opportunities. Non-market systems: credit coops, ROSCAs (rotating savings and credit associations), informal credit/insurance, sharecropping. Non-market does better because: better monitoring of neighbors, better enforcement (relationships matter, courts nonexistant, laws not enforced). group loans improve monitoring(neighbors)/enforcement(everyone pays otherwise), but can incentivize group default (doesn't in practice). credit coops (MIT FCU) are large-scale versions, but matters less today in developed world. ROSCAs: pool money, rotate large disbursements---pay for large, indivisible items (primitive saving)---widespread. (Besley---nonmarket institutions for credit/risk sharing.)

Lecture 17: Poverty trap (Banerjee + Newman): two countries same except for credit constraints diverge economically.  two countries same except for inequality will diverge economically. in both, poverty begets poverty. two people w/ $400 total. machine + person costs $200, gets $1000 profit.  person alone gets $10 profit. equal, unconstrained->both buy machine, profit = $1600. equal, constrained->no need to borrow, profit=$1600. unequal, no constraints->poor borrows, $1600 though poor may pay interest to rich. unequal, constrained, poor can't get machine, so profit=$810. implications: underdevelopment persists w/o credit changing, and colonialism has long effect. de Mel + McKenzie + Woodruff: randomly give cash/equipment to Sri Lankan businesses, estimate return on capital: 60%/year, but loans have 12-18% interest. Implies credit constraint, since loans make sense, but people _can't_ get them. effects of MFI (microfinance institutions): Banerjee, Duflo, Glennsterner, Kinnan: test microcredit causing economic/social improvement (there's already correlation). randomized eval (treat 52/104 villages) of spandana (MFI like Grameen bank, no empowerment/training tracks): small group loans to nonmigrant adult women. previous borrowing was 69% of people, not through MFI: moneylender, family, friends. 31% had small household business. consumption smoothing: no health insurance, 25% life insurance, 34% savings account. spandana treatment group got more MFI loans. 30% used to start new businesses, more to durable than non-durable (forward-thinking).  no effect on child education, women empowerment, or health 12-18 months after loan, but that's short time horizon.

Lecture 18: Poor save/trade free credit to smooth consumption. ROSCAs popular to make pretend savings account. Also share checks to get smaller payments more frequently (Collins + Morduch: Daily Grind). Burgess + Pande: study indian rule of open 4 rural banks for every 1 urban on equal terms (offer savings, jobs, credit, advice): see some increase in rural credit, and some decrease in poverty, but hard to say if it was from savings. dupas + robinson: can save way out of credit trap. offered no interest/high withdrawl fee savings accounts: bad account---if people sign up for it, they are credit constrained: 86% said hard to save $ at home. Women saved way more than men. reason: women more present-biased than men, but sophisticatedly so: high withdrawl fees commit to savings. present biased/time inconsistent/hyperbolic discounter = push off event to tomorrow, but don't care about 1-day different a month from now. naive=always push to tomorrow, sophisticated=commit to what you said a while back. Ashraf, Karlan, Yin: SEED (savings account) offers interest, target date for withdrawl, specify savings goal. 3000 clients: half offered education/goals + SEED, quarter was pure control, quarter got education but no SEED. asked questions about money now vs. later to see if itme inconsistent. turns out time inconsistent people want SEED most---sophisticated.

Lecture 19: Morduch: microinsurance survey: why no microinsurance for poor? assymetry of information---moral hazard (insurance makes you take more risks) + adverse selsection (riskier people want coverage more), also overhead at small scale. Townsend: villages gain if pool insurance money (but it doesn't happen in practice). life insurance very uncommon due to assymetry but people have burial societies to eliminate those. health insurance is same way, so need way to reduce big risks and cover common things to provide smoothing (makes deductables/premiums hard). rainfall insurance good as there's no assymetry, but need to reinsure against larger correlated failure: too expensive. Udry: credit used as insurance---borrow when bad, loan when good.

Lecture 20: Institutions: rules + social norms that guide interactions. good institutions: enforce property rights, constrain executive, formalize legal system. Film suggested that property rights are not good when colonial-era landowners prevent poor from growth. Mooya + Cloete: land titling (right to sell/use/bequeath) to legitimize urban poor property ownership. theoretical benefits: promote investment in property, use as collateral (credit), allow trade of land, less time protecting property, taxable asset. Evidence: increases value, less time protecting, but poor don't use as collateral much---want to keep. sometimes improves trade when not a beaurocratic mess, other times better left w/o govt. corruption. Besley found (Ghana during rights transition) that property right matter a little (planting trees), but not large ammt., but for given farmer, will invest more in land w/ rights than land w/o. Field finds titling program in peru leads treatment group (already received title) spends more time working, less time protecting property, less child labor.

Lecture 21: (Acemoglu, Johnson, Robinson).  Proximate vs. fundamental causes. Which is fundamental cause of development: geography or institutions (of course they are interwoven). Geography matters: affects humans, disease, plants, technologies used, migration (eurasia---move latitudinally in same climate, america---can't move longitudinally, different climates). South vs. North Korea isn't geography, it's institution. Geography is constant, so if it's fundamental, development should be constant, but it's not (reversal of fortune). Urbanization today positively correlates w/ gdp/capita today, but urbanization/pop. density in 1500 negatively correlates w/ gdp today (reversal of fortune). Doesn't yet mean institution is fundamental, geography not. Industrial revolution was big part of reversal, which _is_ tied to institutions. Colonization left better institutions in settled countries (australia, usa) than extracted/unsettled countries (latin america, middle east, asia). High settler mortality in colonialism (disease)->low settlement, high extraction (institutions)->low development today. property rights in colonialism correlates w/ growth today, settler mortality does as well. SM->Y only by way of PR, so SM is an IV for PR. Question: does PR affect Y. First stage: SM affects PR. RF: SM affects Y. Ratio RF/First Stage shows causal effect of PR on Y.

Lecture 22: International Financial Institutions (IFIs).  IMF: prevent beggar-thy-neighbor foreign exchange policies, loans to allow fixed rates---loans subject to conditionality (balance budgets, etc.)---macro.  WB: promote spread of capital to reconstructing/developing countries, where private markets won't go---provide micro loans for projects, provide technical assistance---also subject to conditions (accept fees, etc.).  Arguments against: exchange rates not fixed anymore, private capital is free-flowing, WB shouldn't try soft issues (gender equality, environment), SAL conditions aren't always good policy, don't want to strengthen bad governments. (Kreuger on WB/IMF).

Lecture 23: Trade has winners and losers.  Banerjee explains that while original theories said low-paid workers and consumers in high-cost countries are aided by trade, this is not the case because developing nations end up being used for high-quality stuff, further widening inequality.  So how to we analyze inequality, and should developing countries just put up barriers? Two explanations of trade. Comparative advantage: if two products and two countries, even if A better than B at making both, A should make all of first product and none of second with one unit of labor, and trade allows both countries to have both products: implies very different countries trade different commodities. Increasing returns to scale (Krugman): Minimum efficient scale of production is larger than market in 1 country, so two similar countries open similar factories, and trade similar products with each other, each benefitting from larger market that can consume efficient scale: implies similar countries trade similar products. Both theories say open trade better than none. Not everyone wins: trade is like technology that converts exports (e.g., textiles) into imports (e.g., wine). But when new technology (e.g., digital cameras) comes out, does everyone win (e.g., Kodak)? No. If more winners (usually consumers) than losers (usally producers), then aggregate GDP grows, and theoreticlaly could see winners compensate losers, though that doesn't happen. May want to protect infant industries (import substitution) so they don't get killed, but identifying them is hard, and you have to open trade of their materials providers, which suffers from chicken/egg problem. Another argument for tariffs: easier than income tax. Note: can't have trade deficit/surplus forever: you'll run out of money or goods to trade, so has to be balanced.

Lecture 24: (Ocampo on latin america structural reforms/foreign investments). Capital flows==money/assets into/out of country, measured in capital account. Current account==goods into/out of country. Like trade, capital account must eventually be balanced (0). Also, current=capital account, as money exchanged for goods. Foreign Direct Investment (FDI): >10% stake, ownership+voting control, usually through subsidiary or M&A. Portfolio investment: purchase of stock/bonds w/o legal control. (S - I) + (T - G) = KA that's kind of profound: if the government is balanced (T-G = 0), and the economy is closed (KA = 0), then we have to save as much as we invest. Benefits of portfolio investment: consumption smoothing: investments can come from outside, competition in banking sector. Costs of PI: hot capital, quickly put in/taken out of country, leading to liquidity runs, affecting other sectors. Indonesia crisis: wages dropped, people worked more and dug into savings. Benefit of FDI: more efficient world resource spending, transfer knowledge worldwide, more competition, exploit intra-firm capital markets (less friction inside firm). Costs of FDI: volatility still high, loss of control, "sweatshop labor"->foreign firms pay more, but still some slack.

Lecture 25: Banerjee---Aid can work, we already have enough proven techniques to have WB just develop those, rather than picking winners in beaurocracy.  Easterly---west has dumped more money into africa than anywhere else, but africa has seen no increase in income since that started. cash transfer is simple aid, but donors usually don't give it: don't trust recipients to do right thing. Easterly poverty trap: Africa is below y=x, so vicious cycle->poverty, whereas we're higher on capacity curve, so virtuous cycle. Transformationalists believe big aid in a few areas can shift curve/change fundamentals enough to kick-start virtuous cycle. Marginalists believe in aid for small-scale interventions to make world better. What to spend money on: Transformationalists have hard time---hard to provably find the 'big thing,' whereas marginalists have found lots of small changes through randomized eval that have positive effect---just pick those w/ low cost/benefit ratio. Will money go to intended use: hard---Kenya example: 100% of money left national govt. for education, median school got 0%, mean was 10-20% (corruption is hard---fix institutions!).