DOC PREVIEW
UIUC ECON 450 - Lecture4-5-6_Econ450

This preview shows page 1-2-3-4-5-34-35-36-37-68-69-70-71-72 out of 72 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 72 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Review of Last Lecture• GDP comparisons across countries and why they might not be accurate• 4 reasons make it hard to compare GDP figures (cost of living differences, under-reporting of income, inaccurate population, negative externalities)• Income mobility of countries over timeProgram Evaluation: Applied EconometricsRichard AkreshDepartment of EconomicsUniversity of Illinois at Urbana-ChampaignIntroduction• How do we figure out which are the most effective programs/policies to implement?• Vocational training for ex-soldiers• Cash transfers for schooling• Micro-credit for small business development• Demobilization and reintegration programs• Community development• Research Question: What is the effect of some treatment T (program/policy) on outcome Y?Problem of Identification• What is the counterfactual? What would have happened to person X’s behavior if he/she had been subjected to an alternative policy T?• Earn more if attended primary school• More likely to be vaccinated if there had been a health clinic in village• Higher income due to vocational training program for post-conflict youth• Improved stability, higher income, more likely to be employed after combatant reintegration program• More formally,Treatment status is denoted:• Ti= 1 if treated• Ti= 0 if not treatedThen each person has a potential outcome:• Yi(1) = person i's outcome if Ti= 1• Yi(0) = person i's outcome if Ti= 0Example: What does access to textbooks do to children’s test scores in school?• Note: Many children in developing countries do not have textbooks at school• Yi(1) average test scores of kids in school i if school i has textbooks (treated)• Yi(0) average test scores of same kids in sameschool i in same year if school i has NO textbooks (not treated)• Treatment effect: Yi(1) - Yi(0)• Problem: • You can never have a school i both with and without textbooks (treated and not treated) at the same time.• We want to know how the treated units would have done in the absence of treatment.How can you solve this?• Can never know the effect of having textbooks on a school i in particular, but can hope to learn the average effect of textbooks on schools:E [Yi(1) - Yi(0)]• If you data on lots of schools, some with textbooks and some without, then take the average test scores in both groups and take the difference between the average test scores in schools with texts and average test scores in schools without texts• Problem with Naïve estimate• Possible differencesWe would like to measure the average treatment effect on the treated (ATT) defined as:• ATT = E [Yi(1) - Yi(0) | Ti=1]= E [Yi(1) | Ti=1] - E [Yi(0) | Ti=1]To clarify terminology: • (Yi(1) | Ti=1)• (Yi(0) | Ti=1)• What is E [Yi(0) | Ti=1] ?• We usually observe the following:D = E [Yi(1) | Ti=1] - E [Yi(0) | Ti= 0]• Add and subtract E [Yi(0) | Ti=1] (this is the counterfactual and does not exist)D = E [Yi(1) | Ti=1] - E [Yi(0) | Ti=1] + E [Yi(0) | Ti=1] - E [Yi(0) | Ti= 0]D = E [Yi(1) - Yi(0) | Ti=1] + {E [Yi(0) | Ti=1] - E [Yi(0) | Ti= 0]}• E [Yi(1) - Yi(0) | Ti=1] • ATT (average treatment effect on the treated)• E [Yi(0) | Ti=1] - E [Yi(0) | Ti= 0]• Selection biasSolution to the Identification Problem: Randomized Evaluations• Sample of N units selected from population• This sample does not have to be random—it just tells you to what group your results generalize• Randomly divide the sample into 2 groups:• Treatment group Nt• Control group Nc• Such that Nt+ Nc= N• Treatment group is treated or exposed to some policy and control group is not treated• If the randomization is done correctly, thenE [Yi(0) | Ti=1] = E [Yi(0) | Ti= 0]• Observe outcome Y and compare the mean outcome across treatment and control groups:D = E [Yi(1) | Ti=1] - E [Yi(0) | Ti= 0]which because of randomization is equal to:D = E [Yi(1) | Ti=1] - E [Yi(0) | Ti=1]• Outcomes of units randomized into treatment group would have been the same as outcomes of units randomized into the control group in the ABSENCE of treatment• So if there was no program giving out textbooks, the test scores of the treatment group and the control group would on average be the same• Thus if there is a program, any differences between the groups can be attributed explicitly to the presence of textbooksPotential Problems with Randomized Experiments• Not always feasible: Cannot be done for macroeconomic policies, Central Bank policy, export/import tariffs• Works well for micro issues• Cost: • Expensive because must interview treatment AND control groups• Often difficult to implement properlyProblems (continued)• Political/Ethical issues: Critical issue that derails lots of randomized program evaluations• Explain benefits of program evaluation: Provides a public good so other countries/NGOs learn which programs to implement. • Some think randomized evaluations should be subsidized.• Randomized rollout due to budget constraints: Often, NGOs can be convinced to do a randomized staggered rollout in which the program is introduced progressively and the late adopters serve as the control group for those who get treated first.• Split sample villages randomly into 3 groups.• Year 1: group 1 receives treatment, groups 2 & 3 are control groups• Year 2: groups 1 & 2 receive treatment, group 3 control group• Year 3: all groups receive treatment, no control group, experiment overProblems (continued)• Currently no incentives for NGO/ government agency to want to learn if their program is successful. If program does not achieve objectives, could lead to loss of jobs for NGO workers, loss of donations• Limited duration of experimentProblems (continued)• Mixing up of treatment and controlsProblems (continued)• Generalizability (geographic or demographic)• General equilibrium effectsProblems (continued)• Hawthorne and John Henry Effects: Treatment and control groups may behave differently if they know they are being observed: • Hawthorne Effect• John Henry EffectPart 2Review of identification problem• Cannot get estimate of program impact for each school—best we can do is get average impact of program on group of schools by comparing them to a similar group of schools who were not exposed to the program• Problem of selection bias• Critical goal of program evaluation: to establish credible comparison group• What’s the challenge?• Average differences between 2


View Full Document

UIUC ECON 450 - Lecture4-5-6_Econ450

Download Lecture4-5-6_Econ450
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Lecture4-5-6_Econ450 and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Lecture4-5-6_Econ450 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?