Introduction: The Making of Modern Development Economics

What can be said about the general trend of absolute poverty since 1981? Where has absolute poverty been increasing?

1. Decreasing

2. Sub-Saharan Africa, Middle East & North Africa

What can be said about the behavior of governments & individuals in developing countries?

• politics are often not benevolent, or ineffective

• individuals do not act rationally

What does modern development economics deal with? What are deviations from optimal behavior of rational agents?

Modern DevEc:

• acknowledge limitations to rational behavior

• takes into account institutional details

Deviations:

• preferences: time inconsistency & limited self-control

• belief formation: inattention

• decision making: social & cultural norms, cognitive load

How are time inconsistencies & limited self-control calculated? When do we have a preference change? Consequently, what is the implied self-control issue?

if ß<1: two future periods deducted from each other are less pronounced than the difference between today & tomorrow (discounting)

→ break in discounting moves together with time

preference change: when tomorrow become today

→ self-control issue:

• savings: subject saves less than they planned

• health, agriculture: less investment than planned

→ not only affect poor people but poverty makes the consequences more severe

How is time inconsistency measured? When do we achieve the point of indifference for each question? What happens if we set the RHS equal? What is the present bias indicated by & under which assumption?

With an incentivized survey experiment:

• receive $10 immediately or a payment of x1> $10 in one month

• receive $10 in 12 months or a payment of x2 > $10 in 13 months

→ ask questions for a menu of increasing values of x1, x2

→ Determine switching points x1_bar,x2_bar between earlier & later payment

Point of indifference q1: u(10) = ß(delta)u(x1_bar)

Point of indifference q2: u(10) = (delta)u(x2_bar)

Right-Hand Sides of the equation: ß = u(x2_bar)/u(x1_bar)

Present bias (ß<1) for subjects with x1_bar>x2_bar → assumption: monotonicity

What can be said about the standard model with respect to belief formation? What can be said about inattention in modern development economics?

• subjects form beliefs using all available information → if new information arrives, beliefs are updated

Modern DevEc:

• individuals are inattentive to important (salient) aspects of a decision

• shifting individuals’ attention is not easy

  • if shifted - effects are often temporary

• applications: inattention to existing options, returns, costs, prices, government policies, etc.

What can be said about the standard model with respect to social & cultural norms? What can be said about them in modern development economics?

Standard model: no role of social & cultural norms

Modern DevEc:

• play key role in shaping individual behavior

• most individuals have a strong preference for behaving in accordance with social norms

• applications: education (role of women/girls), health (sexual behavior, vaccinations), savings (role of women, taxing relatives)

What can be said about the standard model with respect to coginitive load? What can be said about them in modern development economics?

Standard model: zero cost of processing information & making optimal decisions

Modern DevEc:

• poverty in itself is a limitation, as it increases cognitive load

  • the higher the cognitive load the more subjects use System 1 for thinking: fast, automatic, emotional, stereotypical

  • good decisions can often only be made by using System 2: slow, effortful, infrequent, logical, calculating

• Implication: subjects suffer from poverty due to a reduction in the quality of their decision making

What can be said about the politics & implementation of poverty reduction interventions? What is the plumber's approach?

• often ineffective politics of poverty reduction, focusing on large budgets & complex planning without ensuring practical implementation for the poor

Plumber approach: prioritize the practical aspects of policy implementation over theoretical policy design

What is the Credibility revolution?

1980s: hardly anyone took anyone else’s data analysis seriously

2010: credibility revolution has fundamentally changed applied econometrics

• shift towards a focus on research design to identify causal effects (moving away from naive regression analysis)

• Focus on research design:

  • emphasis on clear identification of causal effects using exogenous variation with experimental designs becoming the gold standard

What are things that need to be considered before an intervention is implemented at large scale?

• define an outcome of interest (e.g. health status)

• look for a specific intervention or “program”, that might improve the outcome of interest (e.g. deworming program)

• produce evidence on how effective the program is (impact evaluation)

• decide whether it is worth to scale up the program (cost-benefit analysis)

• inform the decision makers (political, or implementation stage)

What is the evaluation problem?

• comparison over time is insufficient → reason: other factors also change

• impossible to evaluate the impact for each individual → average impact of the program on a group of individuals

  • compare treated to non-treated individuals

What is the issue with a naive evaluation strategy?

differences between the treated & control group are attributable to:

• pre-existing differences (selection bias), or

• impact of the program

What is the golden standard for eliminating selection bias? What can be said about the difference between control & treatment in a large sample? What can the statistically significant difference in the outcome between groups be interpreted as?

Randomization

large sample: those exposed to the program and those who are not will not differ on average

Significant difference: causal effect of the treatment

What does stratification help with? How do you test if the characteristics between control & treatment group are balanced?

• Improve the balancedness of treatment & control group

Balancedness check:

• F-test: regress the treatment indicator on the characteristics & run F-test for joint significance of characteristics in predicting the treatment assignment

  • balancedness is given if you cannot reject the hypothesis that the characteristics do not jointly predict the treatment assignment

• t-tests: for each characteristic, run a t-test for the equality of means between treatment & control

  • the treatment assignment is balanced in the respective characteristic if you cannot reject the hypothesis that the means are equal

What is ITT used for? How is it applied?

Intention-to-Treat: Selection bias (e.g. parents move children from a class without the program to a school with the program)

Application:

• ignore selection & compare outcomes between initially assigned groups (always causal; reason: initial assignment was random)

  • we know that we underestimate the true effect relative to a situation without selection

    • We know that the treatment effect on the treated is larger than the ITT

What are caveats of randomized evaluations?

• attrition bias (dropping of subjects from sample)

  • differential attrition: those who participate in the program have different attrition rates than those who do not (e.g. school treatment reduces drop-out rates among treated students only)

    • → the sample of students in the treated schools (or classes) differs from students in the control group

    • If students who are prevented from dropping out are the weakest, the comparison of achievement may be biased downwards

• Spillovers: if treatment affects untreated subjects

  • partial solution: randomization at the level of groups (rather than individuals)