AfricaAidDeconfoundingAnalysis_ConlinThesis Dataset

Overview

AfricaAidDeconfoundingAnalysis_ConlinThesis is a composite dataset of spatially and temporally rich inputs compiled to support Cindy Conlin’s thesis on the causal impacts of foreign aid in Africa.  Rather than a single table, the repository groups together numerous public data sources that describe where and when foreign aid and other development‑relevant phenomena occur.  These sources include geocoded records of Chinese and World Bank projects, natural‑resource deposits, disaster events, political regimes and socio‑economic indicators.  By combining them, the thesis seeks to control for potential confounding factors when evaluating the effect of aid on outcomes such as economic growth, conflict or accessibility.

The repository is organized by source.  Each sub‑folder contains data files (e.g., CSVs, shapefiles or zipped archives) together with original documentation from the source.  Below is a high‑level summary of the main data sets included.  When using this collection you should cite both this dataset and the original providers listed below.

Main data sources

Data source	Description & key facts	References
AidData Chinese Official Finance (Geocoded)	The geocoded Chinese Official Finance dataset (AidData version 1.1.1) pinpoints 3,485 Chinese Government‑financed projects implemented between 2000 and 2014 across 6,190 locations in 138 countries.  The data capture both Chinese aid and non‑concessional official finance, totalling USD 273.6 billion.  Each record contains project attributes (sector, status, funding amounts), geographic coordinates and classification into ODA‑like, other official finance (OOF) or vague flows.	AidData Research and Evaluation Unit, Geocoded Global Chinese Official Finance Dataset, Version 1.1.1.
AidData World Bank Geocoded Research Release 1.4.2	A geocoded dataset of all World Bank IBRD/IDA projects approved from 1995 to 2014.  It tracks $630 billion in commitments and $389 billion in disbursements for 5,684 projects at 61,243 locations.  The package includes tables of projects, locations and transactions that can be linked via a unique project ID.	AidData, World Bank Geocoded Research Release v1.4.2.
DIADATA – Diamonds Dataset	Provides a comprehensive list of all known diamond deposits worldwide.  Each deposit record includes precise geographic coordinates, the geological form of the diamond and discovery/first production dates.  The data are available as GIS shapefiles and spreadsheets.	Gilmore E. et al., 2005, Conflict Diamonds: A New Dataset (PRIO).
DPI – Database of Political Institutions	Institutional and electoral results data covering about 180 countries from 1975–2020.  It measures checks and balances, government tenure and stability, party ideology and fragmentation.	Inter‑American Development Bank / World Bank, Database of Political Institutions.
EM‑DAT – International Disaster Database	Contains records of over 27,000 mass disasters from 1900 to the present compiled from UN agencies, NGOs and press sources.  EM‑DAT focuses on major disasters meeting at least one of four inclusion criteria (≥10 fatalities, ≥100 affected people, declaration of emergency or call for international assistance).	CRED, EM‑DAT – The International Disaster Database.
EU Global Accessibility Map (Travel time to cities 2015)	Developed by Weiss et al. (2018) using OpenStreetMap and remote‑sensing data, this raster surface quantifies travel time to the nearest city for 2015 at ~1 km resolution.  The authors integrate ten global layers capturing transport networks and terrain; results show that 50.9 % of individuals in low‑income settings live within an hour of a city compared with 90.7 % in high‑income settings.	Weiss D. et al., 2018, “A global map of travel time to cities”.
GEMDATA – Gemstone Deposits	A global dataset of 1,022 gemstone sites in 61 countries.  Each site includes latitude/longitude coordinates, the type of gemstone (e.g., ruby, sapphire, emerald) and discovery and first production dates.  Diamonds are excluded (see DIADATA).	Lujala P., 2009, Deadly Combat over Natural Resources: Gems, Petroleum, Drugs, and the Severity of Armed Civil Conflict.
PETRODATA – Petroleum Deposits	Lists all known oil and gas deposits worldwide, with separate datasets for on‑shore and off‑shore fields.  Deposits are stored as polygons in shapefiles and as spreadsheets containing latitude/longitude coordinates of the centroid.  Each deposit is coded with resource type (oil/gas), discovery and production dates, basin name and source information.	Lujala P., Rød J., & Thieme N., 2007, Fighting over Oil: Introducing a New Dataset.
GEMDATA & PETRODATA supplemental resources (ERRATA, dDIAL, etc.)	The repository includes shapefiles and ancillary tables (e.g., ERRATA.dbf, dDIAL.dbf) that correct or extend the GEMDATA and PETRODATA datasets.	Same as above.
PLAD – Political Leaders’ Affiliation Database	A manually compiled dataset that provides geocoded information on the birthplace and ethnicity of political leaders, along with tenure, education and other attributes.  Data are collected from encyclopedias and news reports; the codebook and data are available from Göttingen University.	Bomprezzi P. et al., 2025, Wedded to Prosperity? Spousal Favoritism in Foreign Aid and Regional Development.
Polity5 Dataset (Polity V)	Codes the authority characteristics of states worldwide from 1800–2018 and classifies regimes along a spectrum from autocracy to democracy.  Polity scores range from −10 (hereditary monarchy) to +10 (consolidated democracy).  The dataset records components of executive recruitment, constraints on executive authority and political competition.	Center for Systemic Peace, Polity Project.
UCDP Georeferenced Event Dataset (GED)	Produced by the Uppsala Conflict Data Program (UCDP), this dataset records individual events of organized violence worldwide.  It covers annual and disaggregated events from 1989–2024 with dates resolved to single days and locations down to villages.  UCDP is the main provider of data on organized violence.	Davies S. et al., 2025, Organized violence 1989–2024 (UCDP).
World Development Indicators (WDI)	The World Bank’s flagship indicators dataset.  It provides annual socio‑economic indicators (e.g., GDP, population, education, health, infrastructure) for most countries.	World Bank, World Development Indicators (various years).
WorldPop	Gridded population counts and demographic indicators produced by the University of Southampton; provides high‑resolution population estimates for every year.	WorldPop project.
country_regions, all_iso_codes and DevFi classification	Support files used for geographic harmonisation.  all_iso_codes.csv lists ISO‑3166 country codes with their regions (sourced from Luke’s GitHub).  DevFi_Classification.xlsx contains the OECD Development Finance codelist downloaded on 25 May 2023.  dhs_clusters.csv contains DHS cluster data provided by the AI & Global Development Lab.	See remarks in all_iso_codes_readme.txt.
Other folders (Dreher, GDIS, GEE, GOLDDATA, OECD, UCDP, etc.)	These folders contain supplementary datasets such as econometric covariates, natural resource data and classification codes.  Users should consult the source documentation in each folder for details.	Varies by source.

Usage

The repository is hosted on Hugging Face.  To load it with the Python datasets library you can specify the dataset name and select the files you need.  For example, to access the geocoded AidData Chinese finance tables:

from datasets import load_dataset
# load the all_flow_classes table
china_finance = load_dataset(
    "cjerzak/AfricaAidDeconfoundingAnalysis_ConlinThesis",
    data_files={"all_flow_classes": "AiddataChinav1.1.1/GeoCoded_China_Data_Merged_Files/all_flow_classes.csv"},
    split="train",
)
print(china_finance[0])

Because the repository contains heterogeneous file types (CSV, shapefiles, Excel, zipped archives), users will often need to download and process files manually.  The data_files parameter allows you to specify particular files within the repository.  Geospatial files (e.g., .shp, .dbf) can be read with GIS libraries such as geopandas.

License

All files in this dataset are distributed under the BigScience BLOOM RAIL 1.0 license (Responsible AI License).  You may use the data for research and non‑commercial purposes, but you must comply with the conditions specified by the license.  When publishing results based on these data, please cite both this composite dataset and the individual sources listed above.

Citation

If you use this dataset, please cite it as:

Conlin, Cindy (2025). AfricaAidDeconfoundingAnalysis – Causal analysis of foreign aid in Africa (version 1.0).  Composite dataset hosted on Hugging Face.  https://huggingface.co/datasets/cjerzak/AfricaAidDeconfoundingAnalysis_ConlinThesis (accessed [date]).

In addition, cite the underlying sources listed in the table above (AidData, PRIO, CRED, Centre for Systemic Peace, etc.).  See the original codebooks in each folder for further details on data collection and variable definitions. Please cite the associated publication as: 

@article{daoud2026chinese,
  title={Chinese vs. World Bank Development Projects: Insights from Earth Observation and Computer Vision on Wealth Gains in Africa, 2002-2013},
  author={Daoud, Adel and Cindy Conlin and Connor T. Jerzak},
  journal={World Development},
  year={2026},
  volume={202},
  pages={107328},
  publisher={}
}