electricsheepafrica/africa-synth-governance-public-procurement-data-all

⚠️ Synthetic dataset — Parameterized from published SSA literature, not real observations. Not suitable for empirical analysis or policy inference.

African Public Procurement Data

Abstract

A synthetic dataset modeling public procurement contracts across 12 sub-Saharan African countries (2018–2025), parameterized from procurement authority reports, OECD assessments, and anti-corruption research. Contains 10,000 records per scenario across three transparency scenarios (baseline, high_transparency, low_transparency), with 11 variables covering procurement methods, contract values, bidder counts, single-source rates, contract completion, price benchmarking, and award timelines. Designed for ML classification, anomaly detection, and anti-corruption research in the public procurement domain.

1. Introduction

Public procurement accounts for approximately 12-20% of GDP across sub-Saharan African countries, representing one of the largest areas of government spending and corruption risk. Kenya alone reported over 34,000 contracts worth KES 262.8 billion in FY 2023/2024. Single-source (direct) procurement remains prevalent, with rates varying from 8% in high-transparency systems to 30% in weak governance environments.

Key challenges include: limited competition (average 2-5 bidders per tender), contract completion rates below 60% in many countries, price markups of 20-100% above benchmarks, and procurement timelines exceeding 100 days. The adoption of e-procurement systems (Kenya, Rwanda, Zambia) has improved transparency but coverage remains incomplete.

2. Methodology

2.1 Target Population

Contract-level procurement records for 12 sub-Saharan African countries spanning 2018–2025, across 10 sectors and 4 procurement methods.

Countries included: Nigeria, Kenya, South Africa, Ghana, Tanzania, Uganda, Rwanda, Ethiopia, Senegal, Zambia, Mozambique, Cameroon.

2.2 Parameterization Evidence Table

Parameter	Value Used	Source	Year	Note
Kenya contracts FY2023/24	34,000+ / KES 262.8B	Kenya PPRA MAPS	2024	E-procurement coverage
Single source rate (high)	~8%	OECD benchmarks	2024	Open method dominant
Single source rate (low)	~30%	Brookings Nigeria	2024	Direct procurement prevalent
Average bidders (open)	4-6	ZPPA Zambia	2024	Open tender competitive
Contract completion rate	45-75%	Corruption Watch SA	2024	Varies by sector
Price benchmark markup	20-100%	GTI Global PP Dataset	2024	Corruption indicator
Procurement share of GDP	12-20%	World Bank	2023	SSA average

2.3 Scenario Design

Scenario	Description	Single Source Mult	Completion Mult	Bidder Mult
baseline	Current SSA procurement landscape	1.0×	1.0×	1.0×
high_transparency	Countries with e-procurement and reforms	0.5×	1.2×	1.5×
low_transparency	Weak governance, high corruption risk	2.0×	0.7×	0.6×

3. Dataset Description

3.1 Schema

Column	Type	Units	Range	Description
record_id	int	—	1–10,000	Unique record identifier
country	categorical	—	12 countries	Sub-Saharan African country
year	int	year	2018–2025	Procurement year
procurement_method	categorical	—	4 methods	open, selective, limited, direct
sector	categorical	—	10 sectors	Procurement sector
contract_value_usd_millions	float	USD millions	0.01–1000+	Contract value
num_bidders	int	count	1–20	Number of bidders
single_source	boolean	—	true/false	Single-source procurement flag
contract_completed	boolean	—	true/false	Contract completion status
price_benchmark_ratio	float	ratio	0.5–3.0	Actual price / benchmark price
days_to_award	int	days	7–365	Days from tender to award

3.2 Summary Statistics (baseline)

Variable	Mean	SD	Min	Max
contract_value_usd_millions	15.2	45.3	0.01	850
num_bidders	3.3	2.1	1	15
single_source rate	0.29	—	—	—
completion rate	0.59	—	—	—
price_benchmark_ratio	1.02	0.20	0.5	3.0
days_to_award	85	45	7	365

4. Usage

4.1 Loading with HuggingFace datasets

from datasets import load_dataset

ds = load_dataset("electricsheepafrica/african-public-procurement-data")
ds_low = load_dataset("electricsheepafrica/african-public-procurement-data", "low_transparency")

4.2 Regenerating

pip install numpy pandas scipy matplotlib
python generate_dataset.py --scenario baseline --n 10000 --seed 42
python validate_dataset.py

5. Limitations & Ethical Considerations

1. Synthetic data: Not suitable for audit investigations or official reporting.
2. Country-level aggregation: Does not capture subnational procurement variations.
3. Sector simplification: Procurement categories are aggregated into 10 broad sectors.
4. No contract-level detail: Individual contract clauses, amendments, and variations not modeled.
5. Temporal simplification: Does not capture fiscal year-end procurement spikes.

6. References

1. Fazekas et al., Global Contract-level Public Procurement Dataset, 2024.
2. Corruption Watch, Procurement Watch Report 2024.
3. Kenya PPRA, MAPS Assessment Report 2024.
4. Zambia ZPPA, Procurement Statistics Reports.
5. Brookings, Transparency in Procurement in Nigeria, 2024.
6. OECD, Implementing Procurement Recommendation 2020-2024.
7. World Bank, Doing Business Procurement Indicators.
8. Open Contracting Partnership, OCDS Implementation Reports.

Citation

@dataset{esa_procurement_2026,
  title={African Public Procurement Data},
  author={{Electric Sheep Africa}},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/electricsheepafrica/african-public-procurement-data},
  license={CC-BY-4.0}
}

License

CC-BY-4.0