Skip to content

Latest commit

 

History

History
53 lines (39 loc) · 1.89 KB

README.md

File metadata and controls

53 lines (39 loc) · 1.89 KB
Raw

Marginal Treatment Effects

Replication package for:

  1. "Using Instrumental Variables for Inference About Policy Relevant Treatment Parameters" Mogstad, Santos, and Torgovitsky (2018, Econometrica)
  2. "Identification and Extrapolation of Causal Effects with Instrumental Variables" Mogstad and Torgovitsky (2018, Annual Review of Economics)
  3. "Policy Evaluation with Multiple Instrumental Variables" Mostad, Torgovitsky, and Walters (Forthcoming, Journal of Econometrics)

To load/install dependencies:

using Pkg
Pkg.activate(".")
Pkg.instantiate() # download the correct packages

Then run with

using MarginalTreatmentEffects
menu("/path/to/where/you/want/to/save")

If you want to try to automatically call pdflatex to compile the TikZ figures, do this:

menu("/path/to/where/you/want/to/save", compile = true)

Disclaimer

The solutions to the linear programs are typically not unique and can depend intricately on both software versions and hardware. As a consequence, figures that graph optimal MTR functions (such as Figures 2--7 of MST) may not reproduce, and might reproduce differently on different machines. The bounds on the target parameters are unique, however, and should reproduce on different machines.

Errata

MST (2018, Econometrica)

  • Figure 4: The IV Slope entries in the legend correctly uses the support of the instruments:
    • 𝟙(Z = 2) is now 𝟙(Z = 1)
    • 𝟙(Z = 3) is now 𝟙(Z = 2)
  • Figures 5, 6, 7: The entries in the legend correctly use the support of the instruments:
    • 𝟙(Z = 1) is now 𝟙(Z = 0)
    • 𝟙(Z = 2) is now 𝟙(Z = 1)
    • 𝟙(Z = 3) is now 𝟙(Z = 2)
  • Figure 7: The title was changed from "Order 9 polynomial bounds" to "9th degree polynomial bounds".

MT (2018, Annual Review of Economics)

  • Figure 3: The label on the top tick of Figure 3 was changed from $p(1) - p(2)$ to $p(2) - p(1)$.