Augmented Synthetic Control for Multiple Countries

Downloads

Each dataset is available as a labeled Stata .dta and its source file.

⇩ Download all data (ZIP)stata_codebook.do

Run stata_codebook.do in Stata once to attach long-form per-variable notes to the .dta files.

Load directly in code

Every file loads straight from GitHub (raw URLs). Swap the file name to load any dataset.

Stata

* Stata 14+ : `use` reads an https URL directly
global BASE "https://raw.githubusercontent.com/cmg777/starter-academic-v501/master/content/post/r_sc_multi_country/data/"
use "${BASE}synthetic_panel_2country_intuition.dta", clear
describe
notes

Python

!pip install -q pyreadstat
import pandas as pd
BASE = "https://raw.githubusercontent.com/cmg777/starter-academic-v501/master/content/post/r_sc_multi_country/data/"
df = pd.read_stata(BASE + "synthetic_panel_2country_intuition.dta")

# load every dataset at once
files = ["synthetic_panel_2country_intuition", "synthetic_panel_multicountry"]
data = {f: pd.read_stata(BASE + f + ".dta") for f in files}

# pyreadstat (richest metadata) reads LOCAL files -> download first
import pyreadstat, urllib.request
urllib.request.urlretrieve(BASE + "synthetic_panel_2country_intuition.dta", "synthetic_panel_2country_intuition.dta")
df, meta = pyreadstat.read_dta("synthetic_panel_2country_intuition.dta")

Copy and paste this snippet in Google Colab app. https://colab.research.google.com/notebooks/empty.ipynb

R

# R : haven::read_dta auto-downloads an https URL
library(haven)
BASE <- "https://raw.githubusercontent.com/cmg777/starter-academic-v501/master/content/post/r_sc_multi_country/data/"
df <- read_dta(paste0(BASE, "synthetic_panel_2country_intuition.dta"))

Overview & sources

Companion data for a hands-on R tutorial on the Augmented Synthetic Control Method (ASCM) of Ben-Michael, Feller & Rothstein (2021), demonstrated in a multi-country setting with the augsynth package and its three entry points (single_augsynth, multisynth, augsynth_multiout). Both files here are fully synthetic: the true counterfactual and injected effect are known by construction, so every estimate can be graded against ground truth before the method is turned loose on the real Penn-World-Table EMU data (which lives in the post's reference/ folder and is not documented here). The simulated panels validate that each estimator recovers a known effect and show exactly where plain SCM fails and Ridge augmentation rescues it.

Data sources

Cite this data

Please cite this dataset as follows.

APA

Mendez, C. (2026). Augmented Synthetic Control for Multiple Countries: A Tutorial with augsynth [Data set]. https://carlos-mendez.org/post/r_sc_multi_country/

Ben-Michael, E., Feller, A., & Rothstein, J. (2021). The Augmented Synthetic Control Method. Journal of the American Statistical Association, 116(536), 1415–1427. Abadie, A., Diamond, A., & Hainmueller, J. (2010). Synthetic Control Methods for Comparative Case Studies. Journal of the American Statistical Association, 105(490), 493–505. Papaioannou, S. K. (2021). European monetary integration, TFP and productivity convergence. Economics Letters, 199, 109696.

BibTeX

@misc{mendez2026rscmulticountry,
  author       = {Mendez, Carlos},
  title        = {Augmented Synthetic Control for Multiple Countries: A Tutorial with augsynth},
  year         = {2026},
  howpublished = {\url{https://carlos-mendez.org/post/r_sc_multi_country/}},
  note         = {Data set}
}

@article{benmichael2021augmented,
  author  = {Ben-Michael, Eli and Feller, Avi and Rothstein, Jesse},
  title   = {The Augmented Synthetic Control Method},
  journal = {Journal of the American Statistical Association},
  volume  = {116}, number = {536}, pages = {1415--1427}, year = {2021}
}
@article{abadie2010synthetic,
  author  = {Abadie, Alberto and Diamond, Alexis and Hainmueller, Jens},
  title   = {Synthetic Control Methods for Comparative Case Studies},
  journal = {Journal of the American Statistical Association},
  volume  = {105}, number = {490}, pages = {493--505}, year = {2010}
}
@article{papaioannou2021european,
  author  = {Papaioannou, Sotiris K.},
  title   = {European monetary integration, TFP and productivity convergence},
  journal = {Economics Letters},
  volume  = {199}, pages = {109696}, year = {2021}
}

Variable explorer search & filter all 14 variables

Type to filter by name or label, or use the chips to filter by type. Each row shows a mini distribution. Click a header to sort.

Cross-file variable index

Which file each variable appears in (● = present).

Construction & formulas

Synthetic control answers a counterfactual question: among units that were not treated, find the weighted recipe whose pre-treatment path matches the treated unit, and read the post-treatment gap as the effect (the ATT).

• SCM weight problem: choose convex donor weights W minimizing ||X1 − X0·W||_V subject to w_j ≥ 0 and Σ_j w_j = 1 — an interpolation of the donors, never an extrapolation.
• Augmented (bias-corrected) estimator: τ̂_t^aug = (Y_1t − Σ_j w_j Y_jt) − (m̂_t(X_1) − Σ_j w_j m̂_t(X_j)) — the ordinary SCM gap minus what a prognostic model m̂_t (Ridge when progfunc = "ridge") predicts the residual imbalance should be. With a perfect pre-fit the correction vanishes and ASCM equals SCM.

Two-country DGP (synthetic_panel_2country_intuition): a common trend 40 + 1.2·t + 3·sin(2π·t/9) with Gaussian noise; Borealis follows the trend, Atlantia follows the trend plus an injected effect that switches on in 2012 and grows by 1.5 units/year. By construction Borealis is the counterfactual, so the post-2012 gap equals the injected effect.

Multi-country DGP (synthetic_panel_multicountry): a three-latent-factor model Y(0) = μ + L1·f1 + L2·f2 + L3·f3 + noise with a unit fixed effect; the second outcome is 0.6·Y1 + ν + k·f1 + noise. Treated units C01–C04 are each a sparse convex blend of three named donors (a near-perfect synthetic control provably exists); C05 is placed outside the donor hull (loadings beyond every donor) to stress-test the methods. The injected effect on gdp_index is a jump at adoption plus a yearly ramp (post·jump + slope·(year − adopt)), with a correlated 0.6× effect on trade_index.

The datasets

Switch datasets with the tabs. Each shows the full variable dictionary plus a sortable statistics table with mini distributions and data coverage.

Expand all datasets expand to search (Ctrl/⌘+F) or print across all datasets

Known limitations & caveats

• Synthetic data. Both files are fully simulated; results are internally consistent with the calibration but are not empirical evidence about real-world policy effects. The real euro-area data underlying Part 2 is the Penn World Table panel in the post's reference/ folder, which is not documented here.
• C05 is outside the donor hull by design. No convex blend of the donors can reproduce its pre-treatment path, so plain SCM mis-signs its effect; this is intentional, to show where Ridge augmentation earns its keep.
• Ground-truth columns ship with the data. gdp_index_cf / trade_index_cf (the true counterfactuals) and true_effect_gdp / true_effect_trade (the injected effects) are revealed here for grading; real datasets never give you these.
• Significance is a choice. On these simulated panels the injected effects are real, so headlines are significant — but the jackknife and the wild bootstrap can still disagree (the pooled multisynth effect is significant under the jackknife yet not under the bootstrap). Match the inference tool to the estimator and report when they disagree.