Evaluating the Economic Impact of the Aceh Tsunami

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

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

Overview & sources

Companion data for a hands-on Python tutorial that evaluates the long-run economic impact of the 2004 Indian Ocean tsunami on the Indonesian province of Aceh, replicating Heger & Neumayer (2019) on fully synthetic, calibrated panels. The post treats coastal inundation as a quasi-natural experiment and estimates a dynamic four-period difference-in-differences with pyfixest, an event study with diff-diff, a night-lights dose-response, a synthetic control with mlsynth, and Conley spatial-HAC standard errors validated by Moran’s I. Flooded districts lost about 7.9% of output in 2005 but grew 6.3 percentage points per year faster during 2006–08, and the synthetic control places flooded Aceh +18.3% above its no-tsunami counterfactual by 2012. The data-generating process is calibrated so that re-running the paper’s analyses reproduces its findings (signs, significance, approximate magnitudes); it is for teaching the methods, not for drawing new conclusions about Aceh.

Data sources

Cite this data

Please cite this dataset as follows.

APA

Mendez, C. (2026). Bouncing Back Better? Evaluating the Economic Impact of the Aceh Tsunami [Data set]. https://carlos-mendez.org/post/python_did_sc_tsunami/

Heger, M. P., & Neumayer, E. (2019). The impact of the Indian Ocean tsunami on Aceh's long-term economic growth. Journal of Development Economics, 141, 102365. https://doi.org/10.1016/j.jdeveco.2019.06.008

BibTeX

@misc{mendez2026pythondidsctsunami,
  author       = {Mendez, Carlos},
  title        = {Bouncing Back Better? Evaluating the Economic Impact of the Aceh Tsunami},
  year         = {2026},
  howpublished = {\url{https://carlos-mendez.org/post/python_did_sc_tsunami/}},
  note         = {Data set}
}

@article{heger2019impact,
  author  = {Heger, Martin Philipp and Neumayer, Eric},
  title   = {The impact of the Indian Ocean tsunami on Aceh's long-term economic growth},
  journal = {Journal of Development Economics},
  volume  = {141}, pages = {102365}, year = {2019},
  doi     = {10.1016/j.jdeveco.2019.06.008}
}

Variable explorer search & filter all 41 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 disaster is treated as a quasi-natural experiment: coastal geography (not economic prospects) decided which districts the wave flooded. Treatment is split into event-time windows measured against the omitted 2000–02 baseline: pre (2003–04), tsunami (2005), recovery (2006–08), and postrec (2009–12).

• Dynamic DiD (the headline, on district GDP growth): ΔY_it = β1·D_i·1[pre] + β2·D_i·1[2005] + β3·D_i·1[recovery] + β4·D_i·1[post] + α_i + γ_t + ε_it, with district FE α_i and year FE γ_t. The estimand is the ATT; identification rests on parallel trends.
• Night-lights measure (nl_log): NL_ct = log( ∑_n ( DN_nct + 0.001 ) ) — log of total Digital Numbers (DMSP-OLS brightness, 0–63) summed over a kecamatan’s pixels; the tiny 0.001 keeps zeros loggable. nl_growth is its annual log-difference.
• Dose-response: the same DiD with the on/off dummy replaced by continuous share_pop_flooded / share_area_flooded, or by flood-intensity quintiles.
• Synthetic control: choose donor weights w ≥ 0 summing to 1 to minimise the pre-2005 mismatch min_w (X1 − X0·w)′ V (X1 − X0·w); the treated-minus-synthetic gap after 2005 is the estimated effect.
• Conley spatial-HAC standard errors: a sandwich estimator adding serial (a district with itself over time) and spatial (different districts within 100 km in the same year, weight fading linearly to zero) error correlation; the point estimates never move.

Synthetic data-generating process. GDP and night-lights levels are cumulated from a base year using simulated growth series built as district/kecamatan FE + year FE + a treated×period increment + spatial & serial shocks + Gaussian noise, tuned so a fixed-effects DiD recovers the paper’s coefficients column by column (within about 0.005 on the headline cells). Centroids and continuous flood doses are injected so the spatial standard errors and the dose-response behave like the paper’s without moving the point estimates.

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. There is no real data behind this tutorial; the panels are simulated and calibrated to reproduce the paper's findings. They teach the methods — they are not new empirical evidence about Aceh.
• Tiny treatment group. Only 10 flooded Aceh districts are treated, so point estimates are fragile and standard errors wide; the Aceh-only and city (Kota) columns are especially imprecise (2 flooded Kota districts).
• Observational identification. The estimand is the ATT under parallel trends — an assumption, supported by the flat pre-trend and the null neighbour placebo but never proven.
• Clustered treatment needs honest inference. All treated units sit in one corner of Sumatra; residual growth is spatially autocorrelated (Moran's I = +0.065, p = 0.003). Conley spatial-HAC errors roughly double the recovery effect's SE (0.0146 → 0.0244), downgrading it from a spurious 1% to an honest 5%.
• Missing growth rows are intentional. Growth (gdp_growth, gdp_pc_growth, pop_growth, nl_growth) is undefined in 1999 (no prior year) and gdp/pop growth is also missing for Subulussalam over 2003–06 (an administrative boundary change); every estimator simply drops those rows to match the paper's sample sizes.
• Magnitudes can differ slightly. Signs and significance track the paper closely, but synthetic magnitudes can diverge a little; the night-lights quintile and share-of-area scales follow Table 3's (smaller) units rather than the paper's mutually inconsistent Table 4. See Section 11 of the post for the reproduction audit.