The Emigration Frontier: Income Gaps, Population Scale, and Bounded Migration Adjustment

Replication Package for "The Emigration Frontier: Income Gaps, Population Scale, and Bounded Migration Adjustment"
This dataset provides the full replication package for "The Emigration Frontier: Income Gaps, Population Scale, and Bounded Migration Adjustment" (Mullings, 2026). Drawing on a global panel of 183 countries since 1970, the paper asks why some countries relieve income gaps through large-scale emigration while others do not, despite facing similar incentives to migrate. The central object of analysis is an upper envelope on cross-country emigration outcomes — the "emigration frontier" — populated mainly by small, socially globalized economies with concentrated diaspora networks. Average emigration responses to income gaps are weak and unstable, but upper-tail responses near the frontier are large, persistent, and economically meaningful; reductions in mobility barriers shift this frontier outward asymmetrically, with treatment effects increasing monotonically across quantiles. The empirical strategy combines unconditional quantile regression with a quantile difference-in-differences design exploiting the 1998 opening of European Union accession negotiations with the A8 countries.
The package contains the processed analysis dataset (merged_df_final_revision.csv, 9,667 country-year observations covering 1970–2022) and all code required to reproduce every table and figure in the manuscript and appendices. Code is provided in three languages corresponding to the paper's analytical pipeline: R (Manual_Data_Analysis.R) for upstream data assembly from seven raw public-domain sources; Stata (estimate_final_v3.do) for the main quantile regressions reported in Tables 1–4; and Python scripts 01–05 for the Section 4 distributional figures, the Section 7 difference-in-differences and event-study results (Tables 5–6, Figures 7–10), and Appendices B, D, and E. A driver script (run_all.py) executes the Python pipeline end-to-end. See README.md for software requirements, expected runtimes, and the precise mapping from each output file to its location in the manuscript.
Seven raw input files are not redistributed because they remain subject to the licences of their original providers, but all are openly accessible: Standaert-Rayp bilateral migration (Mendeley Data, DOI 10.17632/cpt3nh6jct.2), Penn World Tables v11.0, KOF Globalisation Index 2025, Barro-Lee education v3, UCDP/PRIO Armed Conflict v25.1, UN World Population Prospects age data, and Quality of Government Standard TS (January 2026). Direct download URLs and expected filenames are documented in README.md. Replicators wishing to regenerate the processed CSV from raw sources should place these seven files in the working directory before running the R script; replicators content with the processed CSV can skip directly to the Stata and Python analyses.