1 / 64

Randomized Evaluations: Applications

Randomized Evaluations: Applications. Kenny Ajayi September 22, 2008. Global Poverty and Impact Evaluation. Review. Causation and Correlation (5 factors) X → Y Y → X X → Y and Y → X Z → X and Z → Y X ‽ Y . Review. Causation and Correlation (5 factors) X → Y causality

shiro
Download Presentation

Randomized Evaluations: Applications

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Randomized Evaluations:Applications Kenny Ajayi September 22, 2008 Global Poverty and Impact Evaluation

  2. Review • Causation and Correlation (5 factors) • X → Y • Y → X • X → Y and Y → X • Z → X and Z → Y • X ‽ Y

  3. Review • Causation and Correlation (5 factors) • X → Y causality • Y → X reverse causality • X → Y and Y → X simultaneity • Z → X and Z → Y omitted variables / confounding • X ‽ Y spurious correlation • Reduced-form relationship between X and Y

  4. Review Exogenous

  5. Review Exogenous Endogenous

  6. Review Exogenous Endogenous External Validity

  7. Review Exogenous Endogenous External Validity Natural Experiment

  8. Practical Applications • Noncompliance • Cost Benefit Analysis • Theory and Mechanisms

  9. Kling, Liebman and Katz (2007) • Do neighborhoods affect their residents?

  10. Do neighborhoods affect their residents? • Positively

  11. Do neighborhoods affect their residents? • Positively • Social connections, role models, security, community resources

  12. Do neighborhoods affect their residents? • Positively • Social connections, role models, security, community resources • Negatively

  13. Do neighborhoods affect their residents? • Positively • Social connections, role models, security, community resources • Negatively • Discrimination, competition with advantaged peers

  14. Do neighborhoods affect their residents? • Positively • Social connections, role models, security, community resources • Negatively • Discrimination, competition with advantaged peers • Not at all

  15. Do neighborhoods affect their residents? • Positively • Social connections, role models, security, community resources • Negatively • Discrimination, competition with advantaged peers • Not at all • Only family influences, genetic factors, individual human capital investments or broader non-neighborhood environment matter

  16. Do neighborhoods affect their residents? • What is Y?

  17. Do neighborhoods affect their residents? • What is Y? • Adults • Self sufficiency • Physical & mental health • Youth • Education (reading/math test scores) • Physical & mental health • Risky behavior

  18. Kling, Liebman and Katz (2007) • Do neighborhoods affect their residents? • X → Y causality • Y → X reverse causality • X → Y and Y → X simultaneity • Z → X and Z → Y omitted variables / confounding • X ‽ Y spurious correlation

  19. Neighborhood Effects • Natural Experiment

  20. Neighborhood Effects • Natural Experiment • Hurricane Katrina • Can we randomly assign people to live in different neighborhoods?

  21. Neighborhood Effects • Natural Experiment • Hurricane Katrina • Can we randomly assign people to live in different neighborhoods? • Sort of…

  22. Moving to Opportunity • Solicit volunteers to participate • Randomly assign each volunteer an option • Control • Section 8 • Section 8 with restriction (low poverty neighborhood) • Volunteers decide whether to move • Compliers: use voucher • Non-compliers: don’t move

  23. Moving to Opportunity

  24. Non-random Participation • Why is this OK? • Volunteers presumably care about their neighborhood environment, so should be the target when thinking about uptake for the use of housing vouchers BUT • Results might not generalize to other populations with different characteristics

  25. Treatment and Compliance • With perfect compliance: • Pr(X=1│Z=1)=1 • Pr(X=1│Z=0)=0 • With imperfect compliance: • 1>Pr(X=1│Z=1)>Pr(X=1│Z=0)>0  Where X - actual treatment Z - assigned treatment

  26. Some Econometrics • Intention-to-treat (ITT) Effects • Differences between treatment and control group means

  27. Some Econometrics • Intention-to-treat (ITT) Effects • Differences between treatment and control group means ITT = E(Y|Z = 1) – E(Y|Z = 0) Y - outcome Z - assigned treatment

  28. Some Econometrics • Intention-to-treat (ITT) Effects • Writing this relationship in mathematical notation Y = Zπ1 + Wβ1 + ε1 Individual wellbeing (Y) depends on whether or not the individual is assigned to treatment (Z), baseline characteristics (W), and whether or not the individual got lucky (ε) π1- effect of being assigned to treatment group = (effect of actually moving) x (compliance rate)

  29. Some Econometrics • Effect of Treatment on the Treated (TOT) • Differences between treatment and control group means Differences in compliance for treatment and control groups

  30. Some Econometrics • Effect of Treatment on the Treated (TOT) • Differences between treatment and control group means Differences in compliance for treatment and control groups • TOT = “Wald Estimator” = E(Y|Z = 1) – E(Y|Z = 0) E(X|Z = 1) – E(X|Z = 0) Y - outcome Z - assigned treatment X - actual treatment (i.e. compliance)

  31. Some More Econometrics • Estimate TOT using instrumental variables • Use offer of a MTO voucher as an instrument for MTO voucher use

  32. Some More Econometrics • Estimate TOT using instrumental variables • Use offer of a MTO voucher as an instrument for MTO voucher use. In mathematical notation: (1) X = Zρd + Wβd + εd • Predict whether people use voucher (X), given their assigned treatment (Z) and observable characteristics (W)

  33. Some More Econometrics • Estimate TOT using instrumental variables • Use offer of a MTO voucher as an instrument for MTO voucher use. In mathematical notation: (1) X = Zρd + Wβd + εd (2)Y = Xγ2 + Wβ2 + ε2 • Predict whether people use voucher (X), given their assigned treatment (Z) and observable characteristics (W) • Estimate how much individual wellbeing (Y) depends on predicted compliance (X) and observable characteristics (W)

  34. Some Econometrics • Effect of Treatment on the Treated (TOT) • Writing the reduced-form relationship in mathematical notation Y = Xγ2 + Wβ2 + ε2 Individual wellbeing (Y) depends on whether or not the individual complies with treatment (X), baseline characteristics (W), and whether or not the individual got lucky (ε) Note:γ2-effect of actuallymoving (i.e. compliance) = effect of being assigned to treatment group compliance rate

  35. Moving to Opportunity

  36. Moving to Opportunity INTENTION TO TREAT

  37. Moving to Opportunity EFFECT OF TREATMENT ON THE TREATED

  38. Results • Positive effects • Teenage girls • Adult mental health • Negative effects • Teenage boys • No effects • Adult economic self-sufficiency & physical health • Younger children • Increasing effects for lower poverty rates

  39. Limitations • Can’t fully separate relocation effect from neighborhood effect • Can’t observe spillovers into receiving neighborhoods

  40. Some Thoughts • We can still estimate impacts in cases where we don’t have full compliance • Policies sometimes target people who are different from the general population (i.e. those more likely to take up a program) BUT • We have to be careful about generalizing these results for other populations (external validity)

  41. Practical Applications Noncompliance Cost Benefit Analysis Theory and Mechanisms

  42. Duflo, Kremer & Robinson (2008) • Why don’t Kenyan farmers use fertilizer?

  43. Duflo, Kremer & Robinson (2008) • Why don’t Kenyan farmers use fertilizer? • Low returns • High cost • Preference for traditional farming methods • Lack of information

  44. Is Fertilizer Use Profitable? • What is Y (outcome)? • Crop yield • Rate of return over the season • What is X (observed)? • Fertilizer • Hybrid seed

  45. Is Fertilizer Use Profitable? • What is Z (unobserved)?

  46. Is Fertilizer Use Profitable? • What is Z (unobserved)? • Farmer experience/ability • Farmer resources • Weather • Soil quality • Market conditions

  47. Correlation or Causation? Fertilizer Use (X) and Crop Yield (Y)

  48. Correlation or Causation? Fertilizer Use (X) and Crop Yield (Y) • X → Y causality • Y → X reverse causality • X → Y and Y → X simultaneity • Z → X and Z → Y omitted variables / confounding • X ‽ Y spurious correlation

  49. Fertilizer Experiment • How could we test the effect of fertilizer use on crop yield using randomization?

  50. Fertilizer Experiment • Randomly select participating farmers • Measure 3 adjacent equal-sized plots on each farm • Randomly assign farming package for each plot • Top dressing fertilizer (T1) • Fertilizer and hybrid seed (T2) • Traditional seed (C) • Weigh each plot’s yield at harvest

More Related