Micro for Development - Risk and development

Question 1

risk and welfare

Answer 1

Question 2

How much money would poor be willing to pay in order to eliminate risks (m)?

Answer 2

Question 3

income = consumption?

Answer 3

• With perfect insurance markets, consumption will only depend on time preference and expected incomes, not on variation
• Options to insure consumption
  
  • Save, borrow
  • Purchase insurance
  • Informal insurance (family , village)
• Coping strategies may be very costly
  
  • Cut some expenditures to safeguard others (dropout school for instance)
  • Search for additional incomes

Question 4

if within village insurance exists, idiosyncratic or covariate shocks matter more?

Answer 4

we would expect idiosyncratic shocks to matter less than covariate shocks (because people could help out each other in case of idiosyncratic shocks as not everyone is affected by them)

Question 5

effects of shocks on consumption - evidence

Answer 5

• Vulnerable are affected more by drought and illness shocks
  
  • Female headed vs male headed
  • Non education versus some education
  • Small versus large landholdings
• Shocks persist
  
  • Drought and illness shocks that occurred more than 2 years ago still affect current consumption

Question 6

why food aid a good shock?

Answer 6

similar across individuals, whereas shocks in other papers are not

Question 7

risk sharing

Answer 7

Question 8

risk sharing and public transfers

Answer 8

• So consumption should not depend on food aid transfer if
  
  • Transfer is indeed a shock conditional on variables that are controlled for.
  • Transfers are not randomly given, so controls are important.
• 1 percent increase in food aid results in 0.08 percent increase in per capita consumption
  
  • So risk sharing rejected
• But is there any risk sharing happening?
  
  • Village aid should affect consumption (conditional on own aid status).
  • Drop the community dummies and replace those by variables which show which shocks the community was subjected to (rainfall).
  • But this could be due to price effects, not risk sharing. Include price index.
  • Individual aid still matters, but village aid not.
• Could measurement error be important?
  
  • Note that community variables is constructed for limited nr of sampled households
  • Measurement error causes a downward reduction in point estimate.
  • Solution IV. This eliminates measurement error.
  • They have two measures of aid in questionnaire, instrument one with the other
• Estimates increase. Village aid matters more than individual aid. Evidence of risk sharing.
• Does food aid crowd out other insurance in village?
  
  • Look at other shocks: Crop shocks. These also affected per capita consumption (when used as a control)
  • Does including whether the village/ individual received aid change this coefficient?
  • Yes, food aid increases the coefficient. There is even less insurance if there is food aid present. Food aid crowds out other insurance mechanisms

Question 9

what do the regression results indicate?

Answer 9

• 1 percent increase in aid per adult yield to a 0.08 percent increase in per capita consumption
• More rainfall, higher yields
• rainfall * dummy for whether it is below the average. Rainfall below the average reduces consumption more than rainfall above the average increases it.
• Village aid seems to have a positive effect on consumption, but it could also be that this is a price effect. Extra food has lowered prices for locally, and increased consumption as a result. To correct for this, they include local prices, see table 3

Question 10

what do the regression results indicate?

Answer 10

• …, this lowers the estimate of village aid, and becomes insignificant. Indeed some of the effect is in the price effect which is highly significant. Lowering prices also increase consumption.
• Food aid was calculated by taking average over the sample of households, this has measurement error. Is instrumented by taking same measure, but calvulated on the basis of a diffeen question (Does this meet the IV requirements, not really). Col 2 are IV estimates
• Better crops than normal increase consumption (and the other way around see table 1). If the village receives food aid, that effect is stronger. So, crop shocks are less insured in villages with food aid. Evidence of crowding out
• In the last colum, the village dummies are replaced with village shock variables, but you have the same effect.

Question 11

attenuation bias of OLS

Answer 11

Question 12

conclusions risk sharing

Answer 12

• Shocks do result in consumption changes
• But also some evidence of risk sharing
• So, are the poor underinsured?
  
  • If there is no cost to insurance, and people are risk averse, they will always fully insure
  • Insurance markets may not exist because of adverse selection and moral hazard problems.
  • Government intervention: forced participation eliminates the adverse selection problem

Question 13

effect of unemployment on consumption in US and Indonesia

Answer 13

• But households behavioral response is very different
• Indonesian household staple food consumption also decreases, these are basic, so costly reduction
• Indonesian households withdraw children from school and increase labor supply
  
  • 13 percent point more likely to stop spending on education all together
  • other household members are 17 percentage points more likely to work for wages when the head of household becomes unemployed
• US household smooth in very different ways
  
  • US households use: (1) transfer income (e.g. unemployment insurance) replaces 15 cents of every dollar of lost income, (2) changes in tax burdens replace 26-35 cents per dollar lost, and 3) savings are used to replace the remaining 25-40
• Indonesian use much more costly methods to deal with risk. Social insurance may have greater benefits.

Question 14

welfare gains to social insurance

Answer 14

• Poor household have higher risk aversion. Will be willing to adopt more costly measures to avoid consumption changes.
• So observed fluctuations in consumption could be due to two reasons:
  
  • Cost of adjusting consumption is low (withdraw savings)
  • γ is high, people work hard to avoid drops in consumption

Question 15

effects of weather on profits & risk

Answer 15

• •Measure portfolio.
• •Estimate production as a function of portfolio and weather
• Based on coefficients derive estimate of expected profits and variance of profits of portfolio
• Farm profits affected by weather variability. This component affects household food consumption while other components does not.
• Does production function show a trade-off between risk and expected profits?, mostly
• Farmers in riskier villages adopt safer strategies
  
  • Rich farmers adjust their portifolios less than poor farmers in response to weather risk.
  • Expected profits reduced by weather variability, but less so for the rich
• farm profits effect consumption, but farm profits net of weather effects not –> households are not insured against rain induced risk

Question 16

what do the regression results indicate?

Answer 16

• Draught animals and liquid assets do not abide by the average profit, risk trade-off; these are likely consumption smoothing assets, resulting in a reverse causality

Question 17

what does this graph show?

Answer 17

For rich weather variability translates into more profit variability than for the poor

Question 18

what do the regression results indicate?

Answer 18

Weather variability reduces average profits, but less so for the rich

Question 19

why is insurance take-up so low?

Answer 19

• useful since
  
  • Poor are not fully insured against risk and adopt costly strategies to reduce risk ex-ante
  • Insurance seems a good idea.
  • Rainfall insurance avoid the moral hazard problem
• why take-up so low?
• Existing Index insurance in India
  
  • Pays out if there is too much rain or too low rain. Actuarial value is 46, is low compared to a typical value of 70 percent in the US.
  • Marketed by local NGO that also assists farmers in other way
• Experiment tested different reasons for low take up
  
  • Too expensive?
  • No trust in that insurance will deliver
  • Cash constraints
  • People do not understand product

Question 20

experiment to lift constraints of low take-up

Answer 20

• Households receive visit during which insurance is marketed.
• Endorsement. Ngo, who is trusted, introduces BASIX NGO
• Cash compensation for participating in study – lifts liquidity constraint
• Education about product. How rainfall, as specified in the contract translates into growth of product.
• In a different regions, random prices
  
  • A non transferable coupon providing discounts of 5, 15 and 30 Rp.
  • This allows for an estimation of the demand curve, and estimate what demand would be if prices were at levels as in the US (70 perc actuarial value)
  • Even with highest discounts, still less than half of the hh take the product ​