electricsheepafrica/african-livestock-insurance-data

---
license: cc-by-4.0
language:
- en
task_categories:
- tabular-classification
- tabular-regression
tags:
- insurance
- livestock
- index-insurance
- ibli
- pastoralist
- sub-saharan-africa
- synthetic
- agritech
- climate-risk
pretty_name: African Livestock Insurance Data
size_categories:
- 10K<n<100K
configs:
- config_name: baseline
data_files:
- split: train
path: data/baseline.csv
- config_name: expanded_index_coverage
data_files:
- split: train
path: data/expanded_index_coverage.csv
- config_name: severe_drought
data_files:
- split: train
path: data/severe_drought.csv
---

# African Livestock Insurance Data

A synthetic dataset of 30,000 records (10,000 per scenario) simulating Index-Based Livestock Insurance (IBLI) programs across 12 Sub-Saharan African countries. Calibrated to real-world IBLI parameters from ILRI programs in Kenya and Ethiopia, including NDVI-based triggers, premium rates, and covariate risk reduction estimates.

## Abstract

This dataset provides synthetic micro-level records of livestock insurance enrollment, climate indices, mortality outcomes, and operational metrics for pastoralist communities in Sub-Saharan Africa. Three scenarios — baseline operations, expanded index coverage, and severe drought — enable researchers to evaluate insurance design, basis risk, payout efficiency, and climate resilience interventions. All parameters are grounded in peer-reviewed IBLI literature from ILRI and partner institutions.

## Introduction

Livestock contributes 10–44% of GDP in African countries and is the primary livelihood asset for over 200 million pastoralists in Arid and Semi-Arid Lands (ASALs). Conventional livestock insurance fails in these settings due to moral hazard, adverse selection, and high transaction costs. Index-Based Livestock Insurance (IBLI), pioneered by the International Livestock Research Institute (ILRI) since 2010, uses satellite-derived Normalized Difference Vegetation Index (NDVI) as a proxy for forage availability, triggering payouts when vegetation falls below the 15th percentile of historical norms.

IBLI has demonstrated a 63% reduction in covariate risk for enrolled households (Jensen et al., 2015; Mude et al., 2009). Premium rates are calibrated at 3.25% of insured value for large stock (cattle, camels) and 5.5% for small stock (goats, sheep, poultry). Despite these advances, basis risk — the mismatch between index triggers and actual losses — remains a key challenge, typically ranging from 0.20 to 0.40.

This synthetic dataset enables researchers to:
- Evaluate index insurance design under varying climate conditions
- Quantify basis risk and its determinants
- Model payout efficiency and farmer welfare outcomes
- Compare insurance maturity across countries and regions

## Methodology

### Data Generation

Records are generated using parametric distributions calibrated to IBLI field data:
- **Enrollment**: Log-normal distribution (μ=5.5, σ=1.2) for household counts
- **Livestock holdings**: Poisson distribution (λ=12 animals/household)
- **NDVI anomaly**: Normal distribution with scenario-specific parameters
- **Mortality**: Linear function of NDVI anomaly and forage availability plus noise
- **Payout**: Deterministic trigger logic based on NDVI percentile vs. threshold

### Parameterization Evidence

| Parameter | Value | Source | Evidence |
|-----------|-------|--------|----------|
| NDVI trigger threshold | 15th percentile | Jensen et al. (2015); Mude et al. (2009) | IBLI uses 15th percentile of 10-day NDVI composites as payout trigger in Marsabit and Wajir counties, Kenya |
| Premium rate (large stock) | 3.25% of insured value | ILRI IBLI program documentation | Actuarially fair rate for cattle and camels based on 30-year mortality records |
| Premium rate (small stock) | 5.50% of insured value | ILRI IBLI program documentation | Higher rate reflects greater mortality volatility in goats and sheep |
| Covariate risk reduction | 63% | Jensen et al. (2015) | RCT in Kenya found IBLI reduced distress sales by 63% during drought |
| Basis risk range | 0.20–0.40 | Chantarat et al. (2013); Carter et al. (2017) | Spatial and temporal basis risk from NDVI-livestock mortality mismatch |
| Payout ratio | 0.50–1.00 | ILRI IBLI payout records | Proportional payout based on severity of index shortfall |
| Livestock unit values | Cattle: $400, Camels: $600, Goats: $60, Sheep: $50, Poultry: $8 | FAO livestock statistics; ILRI field surveys | Regional market prices for ASAL areas |
| Subsidy range | 0–70% | World Bank livestock insurance programs | Government and donor subsidies common in African index insurance |
| Countries covered | 12 SSA countries | FAO pastoralist mapping | Kenya, Ethiopia, Somalia, Tanzania, Uganda, South Africa, Botswana, Namibia, Niger, Mali, Chad, Mozambique |
| Time period | 2010–2025 | IBLI program inception | IBLI launched in Kenya in 2010, Ethiopia in 2012 |

### Scenario Design

| Scenario | Description | Key Parameter Shifts |
|----------|-------------|---------------------|
| **baseline** | Current IBLI operating conditions with standard NDVI index coverage | NDVI μ=0.0, basis risk μ=0.30, mortality base=4%, payout trigger rate=18% |
| **expanded_index_coverage** | Enhanced satellite resolution with multi-index (NDVI + area-yield) approach | NDVI μ=+0.02, basis risk μ=0.22, mortality base=3.5%, payout trigger rate=22%, satisfaction=4.0 |
| **severe_drought** | 2020–2023 Horn of Africa drought conditions | NDVI μ=−0.12, basis risk μ=0.35, mortality base=12%, payout trigger rate=55%, rainfall deficit μ=28% |

## Schema

| Column | Type | Description | Range / Categories |
|--------|------|-------------|-------------------|
| `record_id` | string | Unique identifier (e.g., `BAS-000001`) | — |
| `country` | categorical | Country name | 12 SSA countries |
| `year` | integer | Policy year | 2010–2025 |
| `season` | categorical | Rainfall season | `long_rains`, `short_rains`, `dry` |
| `region_type` | categorical | Agro-ecological zone | `ASAL`, `semi_arid`, `mixed`, `commercial` |
| `livestock_type` | categorical | Primary insured species | `cattle`, `goats`, `sheep`, `camels`, `poultry` |
| `insurance_type` | categorical | Insurance product type | `index_based`, `mortality`, `comprehensive` |
| `enrolled_households` | integer | Number of enrolled households | 5–5,000 |
| `enrolled_animals_count` | integer | Total insured animals | — |
| `premium_per_animal_usd` | float | Premium per animal (USD) | $0.20–$35.00 |
| `total_premium_collected_usd` | float | Total premium for the record | — |
| `ndvi_anomaly_index` | float | Standardized NDVI deviation from mean | −0.6 to +0.6 |
| `rainfall_deficit_pct` | float | Rainfall shortfall from long-term average | 0–80% |
| `forage_availability_index` | float | Composite forage availability (0–1) | 0.05–1.0 |
| `mortality_rate_pct` | float | Observed livestock mortality rate | 0.1–45% |
| `trigger_threshold_pct` | float | NDVI percentile trigger threshold | 8–25% |
| `payout_triggered_flag` | binary | Whether payout was triggered (0/1) | 0, 1 |
| `payout_amount_usd` | float | Total payout amount (USD) | — |
| `payout_ratio` | float | Payout as fraction of insured value | 0–1.0 |
| `basis_risk_score` | float | Index-loss mismatch measure | 0.10–0.55 |
| `claim_settlement_days` | integer | Days from trigger to payout | 3–120 |
| `farmer_satisfaction_score` | float | Farmer satisfaction (1–5 Likert) | 1.0–5.0 |
| `renewal_rate_pct` | float | Policy renewal rate | 10–95% |
| `subsidy_pct` | float | Premium subsidy percentage | 0–70% |
| `distribution_channel` | categorical | Enrollment channel | `insurer`, `NGO`, `cooperative`, `mobile` |
| `climate_resilience_index` | float | Composite resilience measure (0–1) | 0.05–0.95 |
| `livestock_insurance_maturity_class` | categorical | Market maturity stage | `nascent`, `pilot`, `growing`, `mature` |

## Summary Statistics

| Metric | Baseline | Expanded Coverage | Severe Drought |
|--------|----------|-------------------|----------------|
| Records | 10,000 | 10,000 | 10,000 |
| NDVI anomaly (mean) | ~0.00 | ~+0.02 | ~−0.12 |
| Mortality rate (mean) | ~4.0% | ~3.5% | ~12.0% |
| Payout triggered | ~18% | ~22% | ~55% |
| Basis risk (mean) | ~0.30 | ~0.22 | ~0.35 |
| Premium/animal (mean) | ~$8.50 | ~$8.50 | ~$8.50 |
| Satisfaction (mean) | ~3.6 | ~4.0 | ~2.8 |
| Renewal rate (mean) | ~62% | ~72% | ~48% |

*Exact values depend on random seed; run `validate_dataset.py` for computed statistics.*

## Validation

The dataset passes a comprehensive validation suite with the following checks:

### Plausibility Checks (18 tests)
1. Record count per scenario (≈10,000)
2. Column completeness (27 required columns)
3. NDVI anomaly range [−0.6, 0.6]
4. Premium rate calibration (3.25–5.50% of unit value)
5. Payout-trigger consistency (triggered → ratio > 0)
6. Payout ratio range [0, 1.0]
7. Basis risk range [0.10, 0.55]
8. Mortality rate range [0.1%, 45%]
9. NDVI-mortality negative correlation
10. Forage-mortality negative correlation
11. Rainfall deficit-forage negative correlation
12. Premium-payout viability (payout ≤ premium)
13. Satisfaction score range [1.0, 5.0]
14. Renewal rate range [10%, 95%]
15. Subsidy range [0%, 70%]
16. Climate resilience range [0.05, 0.95]
17. No duplicate record IDs
18. All 12 countries represented

### Cross-Scenario Monotonicity Tests (8 tests)
- Basis risk: expanded < baseline < drought
- Mortality: expanded < baseline < drought
- Satisfaction: drought < baseline < expanded
- Renewal rate: drought < baseline < expanded
- Payout trigger rate: baseline < expanded < drought
- Climate resilience: drought < baseline < expanded
- NDVI anomaly: drought < baseline < expanded
- Forage availability: drought < baseline < expanded

### Diagnostic Plots
An 8-panel diagnostic figure is generated covering:
1. NDVI anomaly distribution by scenario
2. Mortality rate distribution
3. NDVI vs. mortality scatter with regression
4. Payout ratio distribution (triggered policies)
5. Basis risk boxplot by scenario
6. Premium per animal by livestock type
7. Satisfaction vs. renewal rate scatter
8. Country distribution bar chart

Run validation: `python validate_dataset.py`

## Usage

```python
import pandas as pd

# Load a single scenario
df = pd.read_csv("data/baseline.csv")

# Load combined dataset with scenario column
df = pd.read_csv("data/combined.csv")

# Example: Compare basis risk across scenarios
for scenario in df["scenario"].unique():
sub = df[df["scenario"] == scenario]
print(f"{scenario}: basis_risk = {sub['basis_risk_score'].mean():.3f}")

# Example: NDVI-mortality relationship
from scipy import stats
corr, pval = stats.pearsonr(df["ndvi_anomaly_index"], df["mortality_rate_pct"])
print(f"NDVI-mortality correlation: r={corr:.3f}, p={pval:.6f}")
```

Load with Hugging Face Datasets:
```python
from datasets import load_dataset
ds = load_dataset("electricsheepafrica/african-livestock-insurance-data")
```

## Limitations

1. **Synthetic data**: Records are parametrically generated, not observed from field programs. Statistical properties are calibrated but individual records are not real.
2. **Simplified climate model**: NDVI anomaly is drawn from a normal distribution; real NDVI exhibits temporal autocorrelation and spatial structure not captured here.
3. **No temporal dynamics**: Each record is independent; there is no panel structure tracking the same households across years.
4. **Simplified basis risk**: Basis risk is modeled as a scalar score rather than decomposed into spatial and temporal components.
5. **Country weights**: Country distribution reflects approximate pastoralist population shares but is not precisely calibrated to census data.
6. **Single trigger mechanism**: Only NDVI-based triggers are modeled; actual IBLI programs may use composite indices or area-yield hybrids.

## References

1. Jensen, N. D., Barrett, C. B., & Mude, A. G. (2015). Index insurance and cash transfers: A comparative analysis from Northern Kenya. *Contemporary Economic Policy*, 33(1), 91-108.
2. Mude, A. G., McPeak, J., Ochieng, J., & Kariuki, J. (2009). Index-based livestock insurance: A means to promote resilience in the Horn of Africa. *ILRI Discussion Paper*.
3. Chantarat, S., Mude, A. G., Turvey, C. G., & Barrett, C. B. (2013). Welfare impacts of index insurance in the presence of a poverty trap. *World Development*, 54, 115-130.
4. Carter, M. R., Janzen, S. A., & Michelson, H. (2017). Estimating the returns to index insurance using experimental and observational data. *World Development*, 94, 230-245.
5. ILRI. (2020). Index-Based Livestock Insurance Program: Annual Report. International Livestock Research Institute, Nairobi, Kenya.
6. FAO. (2022). Pastoralism in Africa: Trends and prospects. Food and Agriculture Organization of the United Nations, Rome.
7. World Bank. (2021). Climate Risk and Insurance in Sub-Saharan Africa. Africa Climate Business Plan.

## Citation

```bibtex
@dataset{african_livestock_insurance_2024,
author = {{Electric Sheep Africa}},
title = {African Livestock Insurance Data},
year = {2024},
publisher = {Hugging Face},
url = {https://huggingface.co/datasets/electricsheepafrica/african-livestock-insurance-data},
note = {Synthetic dataset calibrated to IBLI/ILRI parameters}
}
```

## License

This dataset is licensed under [Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/).

You are free to:
- **Share** — copy and redistribute the material in any medium or format
- **Adapt** — remix, transform, and build upon the material for any purpose

Under the following terms:
- **Attribution** — You must give appropriate credit, provide a link to the license, and indicate if changes were made.