Basic Synthetic Control: The Basque Country Case Study

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_basic_synthetic_control/data/"
use "${BASE}source_data.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_basic_synthetic_control/data/"
df = pd.read_stata(BASE + "source_data.dta")

# load every dataset at once
files = ["source_data"]
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 + "source_data.dta", "source_data.dta")
df, meta = pyreadstat.read_dta("source_data.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_basic_synthetic_control/data/"
df <- read_dta(paste0(BASE, "source_data.dta"))

Overview & sources

Companion data for a beginner-friendly R tutorial on the synthetic control method, applied to the classic Basque Country case study of Abadie and Gardeazabal (2003). The panel is the basque dataset bundled with the Synth R package: annual observations from 1955–1997 for 18 Spanish regional units. The outcome is real GDP per capita (gdpcap, 1986 thousands of USD); the remaining columns are the 13 pre-treatment predictors matched by the algorithm — six sectoral production shares, five education levels, an investment-to-GDP ratio, and a 1969 population density. The tutorial frames the analysis as a causal problem (estimand: ATT), builds a “synthetic Basque” from a weighted recipe of the other regions (85% Catalonia, 15% Madrid), and stress-tests the result with Catalonia and in-space placebos.

Data sources

Cite this data

Please cite this dataset as follows.

APA

Mendez, C. (2026). Basic Synthetic Control with R: The Basque Country Case Study [Data set]. https://carlos-mendez.org/post/r_basic_synthetic_control/

Abadie, A., & Gardeazabal, J. (2003). The Economic Costs of Conflict: A Case Study of the Basque Country. American Economic Review, 93(1), 113–132. https://doi.org/10.1257/000282803321455188 Abadie, A., Diamond, A., & Hainmueller, J. (2010). Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of California's Tobacco Control Program. Journal of the American Statistical Association, 105(490), 493–505. https://doi.org/10.1198/jasa.2009.ap08746

BibTeX

@misc{mendez2026rbasicsyntheticcontrol,
  author       = {Mendez, Carlos},
  title        = {Basic Synthetic Control with R: The Basque Country Case Study},
  year         = {2026},
  howpublished = {\url{https://carlos-mendez.org/post/r_basic_synthetic_control/}},
  note         = {Data set}
}

@article{abadie2003economic,
  author  = {Abadie, Alberto and Gardeazabal, Javier},
  title   = {The Economic Costs of Conflict: A Case Study of the Basque Country},
  journal = {American Economic Review},
  volume  = {93}, number = {1}, pages = {113--132}, year = {2003}
}
@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}
}

Variable explorer search & filter all 17 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

The synthetic control method estimates the missing counterfactual GDP path of the treated unit (the Basque Country) as a weighted average of donor (untreated) regions. Let X1 be the treated unit's pre-treatment predictor vector (13 × 1), X0 the donor predictor matrix (13 × 16), and Z1/Z0 the pre-treatment outcomes (annual gdpcap) for the treated unit and donors.

• Donor weights W* (inner problem): minimize the weighted predictor distance ‖X1 − X0·W‖_V = √[(X1 − X0·W)′ V (X1 − X0·W)] over W ≥ 0, Σ w_j = 1 (a convex combination of donors).
• Predictor weights V* (outer problem): minimize the pre-treatment outcome error (Z1 − Z0·W*(V))′(Z1 − Z0·W*(V)) — the diagonal of V dials how much each predictor matters; chosen by cross-validation on pre-1970 GDP.
• ATT (gap): α̂_1t = Y_1t − Σ_j w_j* · Y_jt for t ≥ 1970 — actual Basque GDP minus the synthetic (donor-weighted) GDP, year by year; the headline ATT averages this over 1970–1997.
• Placebo / MSPE ratio: run the same pipeline treating each donor as if treated; rank by the post/pre mean-squared-prediction-error ratio. A trimmed pseudo p-value compares the Basque ratio to the distribution of comparable-fit placebos.

Education predictors are converted to within-region percentage shares and the top two education levels are collapsed before matching (inside the post's prepare_basque() helper); the sectoral predictors are pooled 1961–1969 averages. The raw, untransformed values are stored here.

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

• Predictor columns are sparse by design. Only gdpcap is observed for all 774 rows. The six sectoral shares (~90 non-missing) and five education levels (~108) are pooled period averages/snapshots, and popdens is a single 1969 cross-section (18 rows). The Synth workflow reads these via period-aggregation arguments, so the gaps are expected, not data errors.
• School columns are raw counts, not shares. school_illit, school_prim, school_med, school_high, and school_post_high are stored as raw figures; the post converts them to within-region percentage shares (and collapses school_high + school_post_high) before matching.
• Region 1 is an aggregate. Row group regionno == 1 (“Spain (Espana)”) is the national total and is dropped from the analysis; the 17 autonomous communities are regions 2–18.
• Column names use underscores. The source basque object names sectoral and school variables with dots (e.g. sec.agriculture); they are renamed to underscores here (sec_agriculture) so the variables are valid Stata names. Values are unchanged.
• Real data, single treated unit. This is real regional accounting data, not simulated. With one treated region and a 16-region donor pool, inference rests on placebos rather than standard errors; the discrete pseudo p-values have limited resolution.