Economic Shocks and Civil Conflict at the Regional Level

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

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

Overview & sources

Replication data for a hands-on Stata tutorial that estimates the causal effect of economic shocks on civil conflict by replicating Hodler & Raschky (2014, Economics Letters). The panel covers 96,591 region–year observations from 5,689 subnational administrative regions across African countries over 1994–2010. Conflict is a binary region-year outcome (1+ or 25+ battle deaths from UCDP); nighttime light intensity (DMSP) proxies local economic activity; and lagged log rainfall and the lagged Palmer Drought Severity Index serve as instruments. The endogenous light variable is instrumented because conflict suppresses light (reverse causality) and light measures activity with error (attenuation). OLS returns a near-zero coefficient; 2SLS using both weather instruments yields about −0.30 on conflict (1+ deaths), with first-stage F-statistics above the Stock–Yogo threshold and a Hansen J p-value of 0.93. This is real observational data, not simulated.

Data sources

Cite this data

Please cite this dataset as follows.

APA

Mendez, C. (2026). IV Estimation with Panel Data: Economic Shocks and Civil Conflict [Data set]. https://carlos-mendez.org/post/stata_iv_panel/

Hodler, R., & Raschky, P. A. (2014). Economic shocks and civil conflict at the regional level. Economics Letters, 124(3), 530–533.

BibTeX

@misc{mendez2026stataivpanel,
  author       = {Mendez, Carlos},
  title        = {IV Estimation with Panel Data: Economic Shocks and Civil Conflict},
  year         = {2026},
  howpublished = {\url{https://carlos-mendez.org/post/stata_iv_panel/}},
  note         = {Data set}
}

@article{hodler2014economic,
  author  = {Hodler, Roland and Raschky, Paul A.},
  title   = {Economic shocks and civil conflict at the regional level},
  journal = {Economics Letters},
  volume  = {124}, number = {3}, pages = {530--533}, year = {2014}
}

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

The IV design isolates the causal effect of economic activity on conflict using weather as an external shifter. With region fixed effects α_i, region-specific trends β_i·t, and year fixed effects γ_t:

• Structural model (second stage): Conflict_it = α_i + β_i·t + γ_t + δ · Light_(i,t−1) + ε_it — the parameter of interest is δ, the causal effect of economic activity on conflict probability (a Local Average Treatment Effect for weather-driven compliers).
• First stage: Light_(i,t−1) = α̃_i + β̃_i·t + γ̃_t + δ̃ · Weather_(i,t−2) + ε̃_it, where Weather_(i,t−2) is lagged rainfall, lagged PDSI, or both.
• Instrument used for the endogenous regressor: lagged nighttime light llnlight01 is instrumented by lagged log rainfall l2lnrain01 and the lagged Palmer Drought Severity Index l2meanpdsi (weather at t−2).
• Exclusion restriction: weather at t−2 affects conflict at t only through its effect on economic activity (light) at t−1 — no direct path. The two-year lag makes a direct effect implausible; with two instruments the Hansen J test checks joint validity (p = 0.93).
• Detrending: each _dt variable is the residual from regressing the raw series on year within each region — equivalent to region-specific linear time trends. All regressions use the _dt variables.

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

• Real observational data. Outcomes are from UCDP (geocoded battle deaths), lights from DMSP-OLS, weather from rainfall/PDSI products — not simulated. Causal claims rest on the instrument's validity, not on randomization.
• Nighttime lights are a noisy proxy. Light measures economic activity with error; classical measurement error biases OLS toward zero (attenuation), which is why OLS (≈0.001) and 2SLS (≈−0.30) differ by orders of magnitude.
• Exclusion is untestable with one instrument. With a single instrument the exclusion restriction cannot be tested; it is defended by the two-year lag. The over-identified (two-instrument) model lets the Hansen J test probe joint validity.
• Conflict is a rare event. Only 4.6% of region-years have 1+ conflict deaths and 1.4% have 25+; estimates are interpreted relative to these low baselines.
• Within vs. between variation. Most variation in lights and rainfall is between regions; the fixed-effects estimator uses only within-region variation, so strong instruments are essential.
• Detrended (_dt) variables only. The analysis uses the residualized series; the raw series are provided for reference but are not the regression inputs.