1 / 49

Tsunao Okumura, Yokohama National University and Emiko Usui, Wayne State University

Concave-Monotone Treatment Response and Monotone Treatment Selection: With Returns to Schooling Application. Tsunao Okumura, Yokohama National University and Emiko Usui, Wayne State University. Introduction.

Download Presentation

Tsunao Okumura, Yokohama National University and Emiko Usui, Wayne State University

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. Concave-Monotone Treatment Response and Monotone Treatment Selection:With Returns to Schooling Application Tsunao Okumura, Yokohama National University and Emiko Usui, Wayne State University

  2. Introduction • Identify the sharp bounds on the mean treatment response under concave monotone treatment response (Concave-MTR) and monotone treatment selection (MTS) assumptions • Empirical application to the returns to schooling • Related research: • Manski (1997) MTR/ Concave-MTR • Manski and Pepper (2000) MTR and MTS

  3. MTR: Concave-MTR: MTR + is concave MTS:

  4. Manski’s bounds (under concave-MTR) are too large and Manski and Pepper’s bounds (under MTR-MTS) are still large • Our bounds assume concave-MTR & MTS. • Concave-monotone assumption is often used in economics; Diminishing marginal returns A unique optimal solution

  5. Empirical Application • Estimate the returns to schooling using the NLSY data. • Compare our estimates with the estimates using only the concave-MTR of Manski and the estimates using only MTR and MTS of Manski and Pepper. • Our estimates are much narrower and close to the point estimates from the previous parametric studies.

  6. Methodology Empirically learn and prior information Purpose: learn about

  7. Manski (1997) • MTR: Then, • Concave-MTR: Then,

  8. E.g. : wage (human-capital production) function MTS: persons who choose more schooling have weakly higher mean wage functions than do those who choose less schooling If MTS & MTR, then

  9. Proposition: Let T be ordered. Let for some and . Assume that satisfies the concave-MTR and MTS. Then, for

  10. These bounds are sharp.

  11. y E[y| z = s] E[ y(t) | z = s ] E[y|z = u] u s t E[y(u)|z= s]

  12. y E[y| z = s] E[ y(t) | z = s ] E[y|z = u] u s t E[y(u)|z= s]

  13. y E[y|z= s] E[y(t)| z= s ] E[y|z = u] u s t E[y(u)|z= s]

  14. y E[y| z = s] E[ y(t)| z=s ] E[y(u)|z = s ] E[y | z = u] u s t

  15. y E[y| z = s] E[ y(t)|z= s ] E[y(u)|z = s ] E[y | z = u] u s t

  16. y E[y| z = s] E[ y(t)| z=s ] E[y(u)|z = s ] E[y | z = u] u s t

  17. y E[y| z = s] E[y(t)|z = s ] E[y | z = u] u* u t s t

  18. y E[ y(t)|z= s ] E[y|z= s] E[ y(t)| z = s’ ] A B C D E[y|z = u] u t s’ s t

  19. y E[ y(t)|z= s ] E[y|z= s] E[ y(t)| z = s’ ] A B C D E[y|z = u] u t s’ s t

  20. y E[y| z = s] E[ y(t)|z= s ] E[y(u)|z = s ] E[y | z = u] u s t

  21. y E[y| z = s] E[y(t)|z = s ] E[y | z = u] u* u s t t

  22. y D C B A E[y|z = s] E[ y(t) |z = s ] E[y|z = u] u s s’ t t

  23. y D C B A E[y|z = s] E[ y(t) |z = s ] E[y|z = u] u s s’ t t

  24. y E[y| z = s] E[y(t)|z = s ] Ours Manski,ConcaveMTR E[y|z= t] Manski&Pepper, MTR-MTS E[y|z = u] t u t s

  25. Manski & Pepper MTR-MTS E[y|z= t] Manski ConcaveMTR Ours E[y|z= s] E[y(t)|z = s ] E[y|z = u] t u s t

  26. y E[ y(t)|z= s ] E[y|z= s] A C E[y|z = u] u t s’ s t

  27. y C A E[y(t)|z =s ] E[y| z = s] E[y|z = u] u s s’ t t

  28. Applying these bounds to the LIE:

  29. Applying these bounds to the LIE:

  30. Applying these bounds to the LIE:

  31. Our bounds on

  32. These bounds are sharp, since it is possible to take the concave-MTR and MTS functions of attaining the lower and upper bounds.

  33. y C B A E[y | z = s] E[ y(t) | z = s ] E[y|z = u] u s s’ t t

  34. y E[y(t)|z = s ] E[y|z= s] A C E[y|z = u] u t s’ s

  35. y E[y(t)|z = s ] E[y|z= s] A C E[y|z = u] u t s’ s

  36. The introduction of the assumption of concavity into MTR-MTS assumptions narrows the width of the bounds on by

  37. The sharp bound on the average treatment effect (Returns to schooling)

  38. Data • The 2000 wave of National Longitudinal Survey of Youth (NLSY) • White male, year-round full-time workers, not self-employed, and the ages 35 -- 45. • Sample size: 1225 individuals • The same data as Manski&Pepper’s (2000), but most recently available data. • t : Schooling years, y : log(wage)

  39. Mean Log(Wages) and Distribution of Schooling: NLSY79, 2000

  40. Upper Bounds on • Card (1999,HLE) surveyed the point estimates on the returns to schooling from the previous studies using parametric methods.

  41. Conclusion • Identify the sharp bounds on the mean treatment response under Concave-MTR and MTS assumptions. • Empirical application to the returns to schooling • Compare our estimates with those using only Concave-MTR (Manski (1997)) and using only MTR-MTS (Manski and Pepper (2000)). • Our bounds are substantially smaller and closer to the point estimates on the existing literature. Future Research • Develop the testing methods using Blundell, Gosling, Ichimura and Meghir (2006) and test the concavity or MTS assumptions.

More Related