Replication paper Internet Use and Understanding Democracy in Africa

Using Afrobarometer Round 6 data (2014), this paper studies how reliance on the Internet and social networks as information sources affects individuals’ understanding of democracy, prioritizing open-ended responses analyzed with ChatGPT to capture conceptual knowledge. Causal identification exploits exogenous variation in Internet access generated by lightning-induced disruptions interacting with 3G coverage. Internet use is associated on average with a negative bias in democratic understanding—especially in high-censorship environments—while it has positive effects in low-censorship contexts, implying that the political impact of Internet diffusion depends critically on media openness.

Replication Steps

Overview

This repository contains all data and code required to reproduce the empirical results, tables, and figures presented in the paper.

The package includes:

• Processed Afrobarometer Round 6 survey data (2014–2015) used in the analysis
• Merged external datasets (V-Dem v13, Polity5, Freedom on the Net, COLDAT, World Bank GDP)
• A master estimation notebook that reproduces all results

Files

Data

• df_survey_afrobarometer.csv – processed survey microdata with constructed variables, instrument measures (lightning × 3G), and merged controls
• V-Dem-CY-Full+Others-v13.csv – democracy and corruption indicators
• p5v2018.xls – Polity5 regime scores
• freedom_of_net.xlsx – internet freedom scores
• COLDAT_dyads.csv – colonial history data
• gdp_data.csv – GDP per capita (constant 2015 USD)

Code

• evaluate/01_estimate.ipynb – master replication notebook (SOS kernel; Python and R).
  Running this file sequentially reproduces all regressions, robustness checks, tables, and figures in the paper.

Requirements

Python (≥3.8): sos-notebook, pandas, numpy, pandas_datareader
R (≥4.0): tidyverse, lfe, readxl, haven, progress

Install SOS Notebook:

pip install sos-notebook
python -m sos_notebook.install

Replication Instructions

1. Download the repository.
2. Ensure all data files are in the same directory structure.
3. Open 01_estimate.ipynb with the SOS kernel.
4. Run all cells from top to bottom.

Data Sources

All datasets originate from publicly available sources (Afrobarometer, V-Dem Institute, Center for Systemic Peace, Freedom House, Harvard Dataverse, World Bank, NASA LIS/VHRFC). Users remain responsible for complying with original data licensing terms.