Replication Code for "The Surrogate Index: Combining Short-Term Proxies to Estimate Long-Term Treatment Effects More Rapidly and Precisely"
This code replicates the application to the GAIN job training program in Athey, Chetty, Imbens and Kang (2019), using a simulated dataset of employment outcomes.
The code can be run directly from Surrogates Metafile.do by setting the global ${surrogates} to point to this repository. If running on MacOS or Linux, change the figure output extension from "wmf" to "png" in "Surrogates Metafile.do".
Alternatively, each file in the folder Code can be run individually. The main .do file is Estimate treatment effects.do, which is contained in Code/Compute Estimates. When the files are run individually, this file should be run first. The other .do files use output from this file to produce figures, tables and scalars that appear in the text. These files can be run in any order.
The data on the GAIN program are from Riverside, CA are from Hotz, Imbens and Klerman (2006). Although the data themselves are not publicly available, a simulated version of the data is saved in the folder Data-raw. The simulation is intended to approximate the main results in the paper, and demonstrate the method used. The file Codebook for GAIN Data.pdf, also in Data-raw, contains further background and variable names.