Replication Package for Do Elite Universities Overpay their Faculty? (Garro-Marin, Kahn, and Lang, 2026)

Quick Readme Guide

Package Overview

This public replication package contains the majority of the code required for replicating Garro-Marín, Kahn, and Lang (2026). The analysis uses restricted-use data from the NCSES Survey of Doctorate Recipients that cannot be published in the package.

Full replication of the paper requires accessing the restricted-use data plus a set code scripts that cannot be published due to the NCSES disclosure rules. See the detailed readme file contained within this package for a full list of the undisclosed files. These files, along with the restricted-use data, can be accessesed through a data-use license.

Folder structure

• code/ contains all the code.
  • Programs in code/build_database do all the data cleaning/
  • Programs in code/data_analysis create all tables and figures in the paper.
• All data is contained in the data/ folder.
  • data/raw/ contains all the raw files required to replicate the dataset.
  • data/additional_processing/ contains intermediate files produced during the execution that are kept for record.
  • data/temporary/ contains intermediate files produced during the execution that are erased afterwards.
  • data/output/ contains all final files used for creating the tables.

Each folder contans a readme.md gives a quick description of the objective and structure of the folder.

Instructions for Replication

1. Edit the working directory and R library path at the top of code/master_do_file.do. The lines that must be edited are appropriately indicated at the top of the code.
2. Edit the working directory and default R package library path in code/R_setup.R. The lines that must be edited are appropriately indicated at the top of the code.
3. Run code/master_do_file.do.

Estimated runtime: Excluding KSS correction, approximately 2 to 3 hours. With the KSS correction it takes 4 to 5 days in the NORC servers.

Advice:

code/master_do_file.do executes replicates all the program and calls three main subroutines:

• code/build_database/master_build.do: cleans all the data.
• code/build_database/correct_KSS_master.do: computes KSS variance correction.
• code/data_analysis/master_tables_and_figures.do: creates all tables and figures.

correct_KSS_master.do is the most compuation- and time-intensive part of the code. It can take up to 5 four days. The results from this correction are just cited in the text and are not required for any of the tables. Replicates can comment out or set run_kss'' to "no" at the top of the master_do_file.do`` to run the replication without this section.

Questions?

Any questions with this package can be referred to César Garro-Marín (cgarrom@ed.ac.uk)