Upload folder using huggingface_hub

README.md

@@ -0,0 +1,207 @@
+---
+pretty_name: African Public Debt Management
+license: cc-by-4.0
+task_categories:
+  - tabular-classification
+  - tabular-regression
+tags:
+  - governance
+  - public-debt
+  - fiscal-policy
+  - sub-saharan-africa
+  - synthetic
+  - lmic
+  - debt-sustainability
+  - macroeconomics
+  - development-economics
+  - debt-management
+  - africa
+  - imf
+  - world-bank
+size_categories:
+  - 10K<n<100K
+configs:
+  - config_name: baseline
+    data_files: data/baseline.csv
+  - config_name: debt_consolidation
+    data_files: data/debt_consolidation.csv
+  - config_name: debt_distress
+    data_files: data/debt_distress.csv
+dataset_info:
+  description: |
+    Synthetic dataset of public debt management indicators for 12 Sub-Saharan African
+    countries across 3 policy scenarios, parameterized from IMF, World Bank, and AfDB
+    reports.
+  features:
+    - name: country
+      dtype: string
+      description: Country name
+    - name: year
+      dtype: int16
+      description: Observation year
+    - name: scenario
+      dtype: string
+      description: Policy scenario (baseline, debt_consolidation, debt_distress)
+    - name: debt_to_gdp_pct
+      dtype: float32
+      description: Total government debt as percentage of GDP
+    - name: external_debt_pct
+      dtype: float32
+      description: Share of total debt owed to external creditors (%)
+    - name: domestic_debt_pct
+      dtype: float32
+      description: Share of total debt held by domestic creditors (%)
+    - name: debt_service_ratio_pct
+      dtype: float32
+      description: Debt service payments as percentage of government revenue
+    - name: concessional_debt_pct
+      dtype: float32
+      description: Share of external debt on concessional terms (%)
+    - name: average_maturity_years
+      dtype: float32
+      description: Weighted average remaining maturity of debt stock (years)
+    - name: average_interest_rate_pct
+      dtype: float32
+      description: Weighted average effective interest rate on debt stock (%)
+    - name: debt_sustainability_rating
+      dtype: int8
+      description: Composite debt sustainability rating (1=low risk, 5=debt distress)
+    - name: refinancing_risk_score
+      dtype: float32
+      description: Refinancing risk score (0-100, higher = greater risk)
+    - name: primary_balance_pct
+      dtype: float32
+      description: Primary fiscal balance as percentage of GDP (positive = surplus)
+    - name: fiscal_space_score
+      dtype: float32
+      description: Fiscal space score (0-100, higher = more fiscal room)
+    - name: gdp_growth_pct
+      dtype: float32
+      description: Real GDP growth rate (%)
+    - name: inflation_pct
+      dtype: float32
+      description: Consumer price inflation (%)
+    - name: reserves_months_imports
+      dtype: float32
+      description: Foreign exchange reserves in months of import cover
+    - name: revenue_to_gdp_pct
+      dtype: float32
+      description: Government revenue as percentage of GDP
+    - name: expenditure_to_gdp_pct
+      dtype: float32
+      description: Government expenditure as percentage of GDP
+---
+
+# African Public Debt Management Dataset
+
+## Overview
+
+This dataset provides synthetic public debt management indicators for **12 Sub-Saharan African (SSA) countries** across **3 policy scenarios** (baseline, debt consolidation, debt distress), totaling **29,988 records** with **19 variables** per record.
+
+The data is parameterized from published indicators and stylized facts reported by the **IMF**, **World Bank**, **AfDB**, and **UNCTAD**, capturing realistic distributions of debt-to-GDP ratios, debt composition, debt service burdens, concessional financing shares, and macroeconomic fundamentals.
+
+## Countries
+
+| Country | Region | Debt Profile |
+|---------|--------|-------------|
+| Nigeria | West Africa | Moderate debt/GDP (~40-55%), high service-to-revenue, Eurobond-dependent |
+| Kenya | East Africa | Elevated debt/GDP (~67%), refinancing pressure, infrastructure-driven |
+| Ghana | West Africa | Post-restructuring (~44%), IMF program, was in debt distress (>80%) |
+| South Africa | Southern Africa | High debt/GDP (~78%), predominantly domestic, SOE risks |
+| DRC | Central Africa | Low debt/GDP (~24%), heavily concessional, IDA-eligible |
+| Ethiopia | East Africa | Low-moderate (~28%), restructuring under G20 Common Framework |
+| Tanzania | East Africa | Moderate (~46%), concessional-heavy, fiscal discipline |
+| Rwanda | East Africa | Rising rapidly (~85%), concessional-heavy, ambitious spending |
+| Mozambique | Southern Africa | Very high (~97%), debt distress, hidden debt legacy |
+| Zambia | Southern Africa | Post-restructuring (~70%), was in debt distress |
+| Senegal | West Africa | Rising (~65%), oil/gas transition, infrastructure investment |
+| Côte d'Ivoire | West Africa | Moderate (~56%), infrastructure-driven, first IDA Eurobond issuer |
+
+## Scenarios
+
+### Baseline
+Current trajectory reflecting prevailing fiscal policies, debt composition, and macroeconomic conditions. Debt-to-GDP ratios average ~58%, with debt service consuming ~25-30% of revenue. Years: 2018–2025.
+
+### Debt Consolidation
+Favorable policy path: fiscal tightening, extended maturities, increased concessional borrowing, lower interest rates, stronger growth. Debt-to-GDP falls ~15%, debt service ratio drops ~25%. Years: 2022–2030.
+
+### Debt Distress
+Adverse scenario: rising debt stocks, refinancing difficulties, loss of concessional access, higher interest rates, weaker growth, rising inflation. Debt-to-GDP rises ~25%, debt service ratio increases ~50%. Years: 2020–2028.
+
+## Variable Descriptions
+
+### Core Debt Indicators
+- **debt_to_gdp_pct**: Total government debt as percentage of GDP. World Bank/IMF threshold of 55% used for low-income countries; many SSA countries exceed this.
+- **external_debt_pct**: Share of total public debt owed to external creditors (multilateral, bilateral, private). SSA average ~42% of total debt stock.
+- **domestic_debt_pct**: Share held by domestic creditors (banks, pension funds, central bank). Complement of external_debt_pct.
+- **debt_service_ratio_pct**: Total debt service (principal + interest) as percentage of government revenue. SSA median ~18.7% for external debt service alone (UNCTAD 2024); many countries exceed 30%.
+
+### Debt Composition & Terms
+- **concessional_debt_pct**: Share of external debt on concessional terms (below-market interest, long grace periods). IDA-eligible countries typically >50%; market-accessing countries <30%.
+- **average_maturity_years**: Weighted average remaining maturity. Shorter maturity = higher rollover risk. Concessional loans: 15-30 years; Eurobonds: 5-10 years.
+- **average_interest_rate_pct**: Weighted average effective rate. Concessional: 0.5-2%; commercial: 5-10%; Eurobonds: 6-12%.
+
+### Risk & Sustainability
+- **debt_sustainability_rating**: Composite rating 1-5 based on debt-to-GDP, debt service ratio, concessionality, and interest rates. 1=low risk, 3=moderate risk, 5=debt distress. Based on IMF-World Bank DSF methodology.
+- **refinancing_risk_score**: Score 0-100 capturing rollover risk from maturity structure, debt level, and concessionality.
+- **fiscal_space_score**: Score 0-100 capturing room for counter-cyclical spending based on primary balance, debt level, revenue mobilization, and debt service burden.
+
+### Fiscal & Macro
+- **primary_balance_pct**: Fiscal balance excluding interest payments (% of GDP). Negative = deficit.
+- **gdp_growth_pct**: Real GDP growth. SSA average ~3.5-4% pre-COVID, recovering post-2020.
+- **inflation_pct**: Consumer price inflation. Varies widely: 3-5% (WAEMU) to 20%+ (Ethiopia, Ghana).
+- **reserves_months_imports**: Foreign exchange reserves in months of import cover. <3 months = vulnerability threshold.
+- **revenue_to_gdp_pct**: Government revenue/GDP. SSA average ~16%, well below OECD ~34%. Critical constraint.
+- **expenditure_to_gdp_pct**: Government expenditure/GDP.
+
+## Data Sources & Parameterization
+
+Parameters derived from:
+
+1. **IMF Regional Economic Outlook: Sub-Saharan Africa** (April 2025) — median debt-to-GDP <60%, debt stabilization trends
+2. **World Bank International Debt Report 2025** — LMIC external debt stocks, service ratios, creditor composition
+3. **AfDB "State of Play of Debt Burden in Africa 2024"** — debt sustainability assessments, liquidity indicators
+4. **UNCTAD external debt sustainability data** — SSA debt service at 18.7% of revenue (3× 2014 levels)
+5. **ONE Data "African Debt"** (Feb 2025) — 21 low-income African countries in/at risk of debt distress
+6. **Debt Service Watch 2024** — SSA spending 55% of revenue on debt service
+
+Key stylized facts embedded in parameters:
+- SSA aggregate debt-to-GDP ~61% (2024), with wide country dispersion (24% DRC to 97% Mozambique)
+- Debt service consuming 15-55% of government revenue across the region
+- Commercial debt share rising from 20% to 43% of total since 2000 (AfDB)
+- 25+ African countries carrying excess debt or high risk (AfDB)
+- Concessional financing declining as countries access capital markets
+
+## Intended Use
+
+- **Debt sustainability analysis** — Train models to predict debt distress risk
+- **Fiscal policy simulation** — Evaluate consolidation vs. expansionary scenarios
+- **Macroeconomic forecasting** — Model debt trajectory under different growth/inflation assumptions
+- **Development finance research** — Study concessional vs. commercial borrowing tradeoffs
+- **ML benchmarking** — Classification (sustainability rating), regression (debt ratios), scenario analysis
+
+## Limitations
+
+- **Synthetic data**: Not a substitute for official national accounts or IMF Article IV data
+- **Cross-sectional stylization**: Each row is an independent draw, not a time series per country
+- **No bilateral/multilateral creditor-level detail**: Concessional share is aggregate
+- **No currency composition**: Exchange rate effects modeled implicitly through scenario shifts
+- **No domestic debt detail**: Treasury bills, bonds, and central bank financing not disaggregated
+
+## Citation
+
+If you use this dataset, please cite:
+
+```
+@dataset{african_public_debt_management_2026,
+  title={African Public Debt Management Synthetic Dataset},
+  author={Electric Sheep Africa},
+  year={2026},
+  url={https://huggingface.co/datasets/electricsheepafrica/african-public-debt-management},
+  license={CC-BY-4.0}
+}
+```
+
+## License
+
+CC BY 4.0 — You may share and adapt with attribution.

data/metadata.json

@@ -0,0 +1,49 @@
+{
+  "name": "African Public Debt Management",
+  "description": "Synthetic dataset of public debt indicators for 12 Sub-Saharan African countries",
+  "n_countries": 12,
+  "countries": [
+    "Cote d'Ivoire",
+    "DRC",
+    "Ethiopia",
+    "Ghana",
+    "Kenya",
+    "Mozambique",
+    "Nigeria",
+    "Rwanda",
+    "Senegal",
+    "South Africa",
+    "Tanzania",
+    "Zambia"
+  ],
+  "scenarios": [
+    "baseline",
+    "debt_consolidation",
+    "debt_distress"
+  ],
+  "n_records_per_scenario": 10000,
+  "n_total_records": 29988,
+  "n_variables": 19,
+  "columns": [
+    "country",
+    "year",
+    "scenario",
+    "debt_to_gdp_pct",
+    "external_debt_pct",
+    "domestic_debt_pct",
+    "debt_service_ratio_pct",
+    "concessional_debt_pct",
+    "average_maturity_years",
+    "average_interest_rate_pct",
+    "debt_sustainability_rating",
+    "refinancing_risk_score",
+    "primary_balance_pct",
+    "fiscal_space_score",
+    "gdp_growth_pct",
+    "inflation_pct",
+    "reserves_months_imports",
+    "revenue_to_gdp_pct",
+    "expenditure_to_gdp_pct"
+  ],
+  "seed": 42
+}

generate_dataset.py

@@ -0,0 +1,463 @@
+"""
+African Public Debt Management Synthetic Dataset Generator
+
+Generates realistic synthetic data for 12 Sub-Saharan African countries across
+3 policy scenarios, parameterized from IMF, World Bank, and AfDB reports.
+
+Sources informing parameters:
+- IMF Regional Economic Outlook: Sub-Saharan Africa (April 2025)
+- World Bank International Debt Report 2025
+- AfDB "State of Play of Debt Burden in Africa 2024"
+- UNCTAD external debt sustainability data
+- ONE Data "African Debt" (Feb 2025)
+"""
+
+import argparse
+import json
+import os
+from pathlib import Path
+
+import numpy as np
+import pandas as pd
+from scipy import stats
+
+SEED = 42
+N_RECORDS_PER_SCENARIO = 10_000
+
+# ── Country parameters (literature-informed) ──────────────────────────────
+# Each dict: debt_to_gdp (mean, std), external_pct (mean, std),
+# debt_service_ratio (mean, std), concessional_pct (mean, std),
+# avg_maturity_years (mean, std), avg_interest_rate (mean, std),
+# primary_balance_pct (mean, std), gdp_growth_pct (mean, std),
+# inflation_pct (mean, std), reserves_months_imports (mean, std),
+# revenue_gdp_pct (mean, std), expenditure_gdp_pct (mean, std)
+#
+# References:
+#   Nigeria ~39-55% debt/GDP, external share rising, high service/revenue
+#   Kenya ~67% debt/GDP, Eurobond-heavy, refinancing pressure
+#   Ghana ~44% post-restructuring (was 80%+), IMF program, debt distress
+#   South Africa ~78-79%, predominantly domestic, SOE contingent liabilities
+#   DRC ~26%, heavily concessional, IDA-eligible
+#   Ethiopia ~28%, restructuring under Common Framework
+#   Tanzania ~46%, moderate debt, strong fiscal discipline
+#   Rwanda ~85%, rising rapidly, concessional-heavy
+#   Mozambique ~97%, debt distress, hidden debt scandal legacy
+#   Zambia ~restructured under Common Framework, ~80% pre-restructuring
+#   Senegal ~55-100% (oil/gas transition), rising
+#   Côte d'Ivoire ~55%, infrastructure-driven borrowing
+
+COUNTRY_PARAMS = {
+    "Nigeria": {
+        "debt_to_gdp": (48.0, 6.0),
+        "external_pct": (42.0, 8.0),
+        "debt_service_ratio": (28.0, 5.0),
+        "concessional_pct": (30.0, 8.0),
+        "avg_maturity_years": (8.5, 2.0),
+        "avg_interest_rate": (7.5, 1.5),
+        "primary_balance_pct": (-2.5, 1.5),
+        "gdp_growth_pct": (3.0, 1.2),
+        "inflation_pct": (18.0, 4.0),
+        "reserves_months_imports": (7.0, 1.5),
+        "revenue_gdp_pct": (7.5, 1.5),
+        "expenditure_gdp_pct": (12.0, 2.0),
+    },
+    "Kenya": {
+        "debt_to_gdp": (68.0, 7.0),
+        "external_pct": (52.0, 8.0),
+        "debt_service_ratio": (32.0, 6.0),
+        "concessional_pct": (35.0, 8.0),
+        "avg_maturity_years": (7.0, 2.0),
+        "avg_interest_rate": (8.0, 1.8),
+        "primary_balance_pct": (-3.5, 1.5),
+        "gdp_growth_pct": (5.0, 1.0),
+        "inflation_pct": (7.0, 2.5),
+        "reserves_months_imports": (4.5, 1.0),
+        "revenue_gdp_pct": (16.0, 2.0),
+        "expenditure_gdp_pct": (22.0, 2.5),
+    },
+    "Ghana": {
+        "debt_to_gdp": (55.0, 10.0),
+        "external_pct": (48.0, 10.0),
+        "debt_service_ratio": (40.0, 8.0),
+        "concessional_pct": (28.0, 10.0),
+        "avg_maturity_years": (6.5, 2.0),
+        "avg_interest_rate": (9.0, 2.0),
+        "primary_balance_pct": (-4.0, 2.0),
+        "gdp_growth_pct": (4.5, 1.5),
+        "inflation_pct": (22.0, 6.0),
+        "reserves_months_imports": (3.0, 1.0),
+        "revenue_gdp_pct": (13.0, 2.0),
+        "expenditure_gdp_pct": (20.0, 3.0),
+    },
+    "South Africa": {
+        "debt_to_gdp": (77.0, 5.0),
+        "external_pct": (12.0, 4.0),
+        "debt_service_ratio": (18.0, 3.0),
+        "concessional_pct": (8.0, 4.0),
+        "avg_maturity_years": (14.0, 3.0),
+        "avg_interest_rate": (7.0, 1.0),
+        "primary_balance_pct": (-4.5, 1.0),
+        "gdp_growth_pct": (1.5, 1.0),
+        "inflation_pct": (5.5, 1.5),
+        "reserves_months_imports": (5.0, 1.0),
+        "revenue_gdp_pct": (27.0, 2.0),
+        "expenditure_gdp_pct": (33.0, 2.5),
+    },
+    "DRC": {
+        "debt_to_gdp": (24.0, 5.0),
+        "external_pct": (75.0, 10.0),
+        "debt_service_ratio": (8.0, 3.0),
+        "concessional_pct": (80.0, 8.0),
+        "avg_maturity_years": (20.0, 4.0),
+        "avg_interest_rate": (2.5, 1.0),
+        "primary_balance_pct": (1.0, 2.0),
+        "gdp_growth_pct": (6.5, 2.0),
+        "inflation_pct": (12.0, 5.0),
+        "reserves_months_imports": (1.5, 0.8),
+        "revenue_gdp_pct": (12.0, 2.5),
+        "expenditure_gdp_pct": (14.0, 3.0),
+    },
+    "Ethiopia": {
+        "debt_to_gdp": (30.0, 5.0),
+        "external_pct": (58.0, 10.0),
+        "debt_service_ratio": (12.0, 4.0),
+        "concessional_pct": (55.0, 10.0),
+        "avg_maturity_years": (15.0, 4.0),
+        "avg_interest_rate": (3.5, 1.5),
+        "primary_balance_pct": (-1.5, 2.0),
+        "gdp_growth_pct": (6.0, 2.0),
+        "inflation_pct": (20.0, 8.0),
+        "reserves_months_imports": (2.0, 0.8),
+        "revenue_gdp_pct": (10.0, 2.0),
+        "expenditure_gdp_pct": (14.0, 2.5),
+    },
+    "Tanzania": {
+        "debt_to_gdp": (42.0, 5.0),
+        "external_pct": (48.0, 8.0),
+        "debt_service_ratio": (15.0, 4.0),
+        "concessional_pct": (55.0, 10.0),
+        "avg_maturity_years": (14.0, 3.0),
+        "avg_interest_rate": (3.5, 1.2),
+        "primary_balance_pct": (-1.0, 1.5),
+        "gdp_growth_pct": (5.5, 1.0),
+        "inflation_pct": (4.5, 2.0),
+        "reserves_months_imports": (5.0, 1.0),
+        "revenue_gdp_pct": (15.0, 2.0),
+        "expenditure_gdp_pct": (18.0, 2.0),
+    },
+    "Rwanda": {
+        "debt_to_gdp": (72.0, 8.0),
+        "external_pct": (62.0, 8.0),
+        "debt_service_ratio": (18.0, 5.0),
+        "concessional_pct": (60.0, 10.0),
+        "avg_maturity_years": (16.0, 3.0),
+        "avg_interest_rate": (3.0, 1.0),
+        "primary_balance_pct": (-5.0, 2.0),
+        "gdp_growth_pct": (7.5, 1.5),
+        "inflation_pct": (7.0, 3.0),
+        "reserves_months_imports": (4.0, 1.0),
+        "revenue_gdp_pct": (16.0, 2.0),
+        "expenditure_gdp_pct": (24.0, 3.0),
+    },
+    "Mozambique": {
+        "debt_to_gdp": (95.0, 8.0),
+        "external_pct": (65.0, 10.0),
+        "debt_service_ratio": (22.0, 5.0),
+        "concessional_pct": (40.0, 12.0),
+        "avg_maturity_years": (10.0, 3.0),
+        "avg_interest_rate": (5.0, 1.5),
+        "primary_balance_pct": (-3.0, 2.0),
+        "gdp_growth_pct": (4.0, 2.0),
+        "inflation_pct": (10.0, 4.0),
+        "reserves_months_imports": (3.5, 1.0),
+        "revenue_gdp_pct": (25.0, 3.0),
+        "expenditure_gdp_pct": (30.0, 3.0),
+    },
+    "Zambia": {
+        "debt_to_gdp": (70.0, 10.0),
+        "external_pct": (58.0, 10.0),
+        "debt_service_ratio": (25.0, 6.0),
+        "concessional_pct": (30.0, 10.0),
+        "avg_maturity_years": (8.0, 2.5),
+        "avg_interest_rate": (7.0, 2.0),
+        "primary_balance_pct": (-3.5, 2.0),
+        "gdp_growth_pct": (4.0, 1.5),
+        "inflation_pct": (12.0, 4.0),
+        "reserves_months_imports": (3.0, 1.0),
+        "revenue_gdp_pct": (18.0, 2.5),
+        "expenditure_gdp_pct": (24.0, 3.0),
+    },
+    "Senegal": {
+        "debt_to_gdp": (65.0, 8.0),
+        "external_pct": (55.0, 8.0),
+        "debt_service_ratio": (16.0, 4.0),
+        "concessional_pct": (45.0, 10.0),
+        "avg_maturity_years": (12.0, 3.0),
+        "avg_interest_rate": (4.5, 1.5),
+        "primary_balance_pct": (-4.5, 2.0),
+        "gdp_growth_pct": (7.0, 2.0),
+        "inflation_pct": (3.0, 1.5),
+        "reserves_months_imports": (5.5, 1.5),
+        "revenue_gdp_pct": (22.0, 2.5),
+        "expenditure_gdp_pct": (28.0, 3.0),
+    },
+    "Cote d'Ivoire": {
+        "debt_to_gdp": (56.0, 6.0),
+        "external_pct": (48.0, 8.0),
+        "debt_service_ratio": (14.0, 4.0),
+        "concessional_pct": (42.0, 10.0),
+        "avg_maturity_years": (11.0, 3.0),
+        "avg_interest_rate": (5.0, 1.5),
+        "primary_balance_pct": (-2.5, 1.5),
+        "gdp_growth_pct": (6.5, 1.5),
+        "inflation_pct": (4.0, 2.0),
+        "reserves_months_imports": (4.5, 1.0),
+        "revenue_gdp_pct": (16.0, 2.0),
+        "expenditure_gdp_pct": (21.0, 2.5),
+    },
+}
+
+# ── Scenario multipliers ───────────────────────────────────────────────────
+SCENARIO_SHIFTS = {
+    "baseline": {
+        "debt_to_gdp": 1.0,
+        "external_pct": 1.0,
+        "debt_service_ratio": 1.0,
+        "concessional_pct": 1.0,
+        "avg_maturity_years": 1.0,
+        "avg_interest_rate": 1.0,
+        "primary_balance_pct": 1.0,
+        "gdp_growth_pct": 1.0,
+        "inflation_pct": 1.0,
+        "reserves_months_imports": 1.0,
+    },
+    "debt_consolidation": {
+        "debt_to_gdp": 0.85,
+        "external_pct": 0.90,
+        "debt_service_ratio": 0.75,
+        "concessional_pct": 1.15,
+        "avg_maturity_years": 1.15,
+        "avg_interest_rate": 0.85,
+        "primary_balance_pct": -0.5,
+        "gdp_growth_pct": 1.20,
+        "inflation_pct": 0.80,
+        "reserves_months_imports": 1.20,
+    },
+    "debt_distress": {
+        "debt_to_gdp": 1.25,
+        "external_pct": 1.15,
+        "debt_service_ratio": 1.50,
+        "concessional_pct": 0.70,
+        "avg_maturity_years": 0.75,
+        "avg_interest_rate": 1.30,
+        "primary_balance_pct": -2.0,
+        "gdp_growth_pct": 0.60,
+        "inflation_pct": 1.50,
+        "reserves_months_imports": 0.70,
+    },
+}
+
+
+def compute_derived_features(df: pd.DataFrame) -> pd.DataFrame:
+    """Compute derived/aggregate features from base variables."""
+    # Domestic debt percentage (complement of external)
+    df["domestic_debt_pct"] = 100.0 - df["external_debt_pct"]
+
+    # Debt sustainability rating (DSR) from debt-to-GDP and service ratio
+    dsr_score = (
+        0.4 * df["debt_to_gdp_pct"] / 100
+        + 0.35 * df["debt_service_ratio_pct"] / 50
+        + 0.15 * (1 - df["concessional_debt_pct"] / 100)
+        + 0.10 * df["average_interest_rate_pct"] / 15
+    )
+    # Map to 1-5 rating (1=low risk, 5=debt distress)
+    df["debt_sustainability_rating"] = np.clip(
+        np.round(1 + 4 * dsr_score / dsr_score.quantile(0.95)).values, 1, 5
+    ).astype(int)
+
+    # Refinancing risk score (0-100)
+    df["refinancing_risk_score"] = np.clip(
+        20
+        + 0.3 * df["debt_to_gdp_pct"]
+        + 0.5 * (100 / df["average_maturity_years"].clip(lower=1))
+        - 0.3 * df["concessional_debt_pct"]
+        + 0.2 * df["average_interest_rate_pct"]
+        + np.random.normal(0, 5, len(df)),
+        0,
+        100,
+    ).round(1)
+
+    # Fiscal space score (0-100, higher = more room)
+    df["fiscal_space_score"] = np.clip(
+        50
+        + 1.5 * df["primary_balance_pct"]
+        - 0.4 * df["debt_to_gdp_pct"]
+        + 0.8 * df["revenue_to_gdp_pct"]
+        - 0.3 * df["debt_service_ratio_pct"]
+        + np.random.normal(0, 6, len(df)),
+        0,
+        100,
+    ).round(1)
+
+    return df
+
+
+def assign_year(scenario: str, rng: np.random.Generator) -> np.ndarray:
+    """Assign observation years with scenario-dependent distributions."""
+    if scenario == "baseline":
+        return rng.choice(range(2018, 2026), size=1, p=[0.06, 0.07, 0.08, 0.10, 0.12, 0.14, 0.18, 0.25])[0]
+    elif scenario == "debt_consolidation":
+        return rng.choice(range(2022, 2031), size=1, p=[0.05, 0.06, 0.08, 0.10, 0.12, 0.14, 0.14, 0.16, 0.15])[0]
+    else:
+        return rng.choice(range(2020, 2029), size=1, p=[0.06, 0.08, 0.10, 0.12, 0.13, 0.14, 0.14, 0.13, 0.10])[0]
+
+
+def generate_scenario(
+    scenario: str, n_records: int, rng: np.random.Generator
+) -> pd.DataFrame:
+    """Generate n_records for a given scenario across all countries."""
+    rows = []
+    records_per_country = n_records // len(COUNTRY_PARAMS)
+    shifts = SCENARIO_SHIFTS[scenario]
+
+    for country, params in COUNTRY_PARAMS.items():
+        for _ in range(records_per_country):
+            year = assign_year(scenario, rng)
+
+            # Apply scenario shifts with random noise
+            dtg = max(5, rng.normal(
+                params["debt_to_gdp"][0] * shifts["debt_to_gdp"],
+                params["debt_to_gdp"][1],
+            ))
+            ext_pct = np.clip(
+                rng.normal(
+                    params["external_pct"][0] * shifts["external_pct"],
+                    params["external_pct"][1],
+                ),
+                5, 95,
+            )
+            dsr = max(1, rng.normal(
+                params["debt_service_ratio"][0] * shifts["debt_service_ratio"],
+                params["debt_service_ratio"][1],
+            ))
+            con_pct = np.clip(
+                rng.normal(
+                    params["concessional_pct"][0] * shifts["concessional_pct"],
+                    params["concessional_pct"][1],
+                ),
+                0, 95,
+            )
+            maturity = np.clip(rng.normal(
+                params["avg_maturity_years"][0] * shifts["avg_maturity_years"],
+                params["avg_maturity_years"][1],
+            ), 1, 40)
+            interest = max(0.1, rng.normal(
+                params["avg_interest_rate"][0] * shifts["avg_interest_rate"],
+                params["avg_interest_rate"][1],
+            ))
+            pb = rng.normal(
+                params["primary_balance_pct"][0] + shifts["primary_balance_pct"] - 1.0,
+                params["primary_balance_pct"][1],
+            )
+            gdp_g = max(-10, rng.normal(
+                params["gdp_growth_pct"][0] * shifts["gdp_growth_pct"],
+                params["gdp_growth_pct"][1],
+            ))
+            inflation = max(0, rng.normal(
+                params["inflation_pct"][0] * shifts["inflation_pct"],
+                params["inflation_pct"][1],
+            ))
+            reserves = max(0.1, rng.normal(
+                params["reserves_months_imports"][0] * shifts["reserves_months_imports"],
+                params["reserves_months_imports"][1],
+            ))
+            rev_gdp = max(1, rng.normal(
+                params["revenue_gdp_pct"][0],
+                params["revenue_gdp_pct"][1],
+            ))
+            exp_gdp = max(rev_gdp, rng.normal(
+                params["expenditure_gdp_pct"][0],
+                params["expenditure_gdp_pct"][1],
+            ))
+
+            rows.append({
+                "country": country,
+                "year": year,
+                "scenario": scenario,
+                "debt_to_gdp_pct": round(dtg, 2),
+                "external_debt_pct": round(ext_pct, 2),
+                "debt_service_ratio_pct": round(dsr, 2),
+                "concessional_debt_pct": round(con_pct, 2),
+                "average_maturity_years": round(maturity, 2),
+                "average_interest_rate_pct": round(interest, 2),
+                "primary_balance_pct": round(pb, 2),
+                "gdp_growth_pct": round(gdp_g, 2),
+                "inflation_pct": round(inflation, 2),
+                "reserves_months_imports": round(reserves, 2),
+                "revenue_to_gdp_pct": round(rev_gdp, 2),
+                "expenditure_to_gdp_pct": round(exp_gdp, 2),
+            })
+
+    df = pd.DataFrame(rows)
+    df = compute_derived_features(df)
+
+    # Reorder columns for clarity
+    col_order = [
+        "country", "year", "scenario",
+        "debt_to_gdp_pct", "external_debt_pct", "domestic_debt_pct",
+        "debt_service_ratio_pct", "concessional_debt_pct",
+        "average_maturity_years", "average_interest_rate_pct",
+        "debt_sustainability_rating", "refinancing_risk_score",
+        "primary_balance_pct", "fiscal_space_score",
+        "gdp_growth_pct", "inflation_pct",
+        "reserves_months_imports", "revenue_to_gdp_pct", "expenditure_to_gdp_pct",
+    ]
+    return df[col_order]
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Generate African Public Debt Management dataset")
+    parser.add_argument("--output-dir", default="data", help="Output directory")
+    parser.add_argument("--records-per-scenario", type=int, default=N_RECORDS_PER_SCENARIO)
+    parser.add_argument("--seed", type=int, default=SEED)
+    args = parser.parse_args()
+
+    os.makedirs(args.output_dir, exist_ok=True)
+    rng = np.random.default_rng(args.seed)
+
+    for scenario in ["baseline", "debt_consolidation", "debt_distress"]:
+        print(f"Generating {scenario} scenario ({args.records_per_scenario} records)...")
+        df = generate_scenario(scenario, args.records_per_scenario, rng)
+        out_path = os.path.join(args.output_dir, f"{scenario}.csv")
+        df.to_csv(out_path, index=False)
+        print(f"  Saved to {out_path} — {len(df)} rows, {len(df.columns)} columns")
+
+    # Save combined dataset
+    dfs = []
+    for scenario in ["baseline", "debt_consolidation", "debt_distress"]:
+        dfs.append(pd.read_csv(os.path.join(args.output_dir, f"{scenario}.csv")))
+    combined = pd.concat(dfs, ignore_index=True)
+    combined_path = os.path.join(args.output_dir, "all_scenarios.csv")
+    combined.to_csv(combined_path, index=False)
+    print(f"Combined dataset saved to {combined_path} — {len(combined)} rows")
+
+    # Save dataset card metadata
+    meta = {
+        "name": "African Public Debt Management",
+        "description": "Synthetic dataset of public debt indicators for 12 Sub-Saharan African countries",
+        "n_countries": len(COUNTRY_PARAMS),
+        "countries": sorted(COUNTRY_PARAMS.keys()),
+        "scenarios": ["baseline", "debt_consolidation", "debt_distress"],
+        "n_records_per_scenario": args.records_per_scenario,
+        "n_total_records": len(combined),
+        "n_variables": len(combined.columns),
+        "columns": list(combined.columns),
+        "seed": args.seed,
+    }
+    with open(os.path.join(args.output_dir, "metadata.json"), "w") as f:
+        json.dump(meta, f, indent=2)
+    print("Metadata saved to data/metadata.json")
+
+
+if __name__ == "__main__":
+    main()

requirements.txt

@@ -0,0 +1,4 @@
+numpy>=1.24
+pandas>=2.0
+scipy>=1.11
+matplotlib>=3.7

validate_dataset.py

@@ -0,0 +1,273 @@
+"""
+African Public Debt Management Dataset Validator
+
+Runs plausibility checks and generates 8-panel diagnostic plots.
+"""
+
+import os
+import sys
+
+import matplotlib.pyplot as plt
+import numpy as np
+import pandas as pd
+from matplotlib.gridspec import GridSpec
+
+OUTPUT_DIR = "data"
+PLOT_DIR = os.path.join(OUTPUT_DIR, "plots")
+
+# ── Plausibility thresholds ────────────────────────────────────────────────
+CHECKS = {
+    "debt_to_gdp_pct": (1, 300),
+    "external_debt_pct": (0, 100),
+    "domestic_debt_pct": (0, 100),
+    "debt_service_ratio_pct": (0, 100),
+    "concessional_debt_pct": (0, 100),
+    "average_maturity_years": (0.5, 40),
+    "average_interest_rate_pct": (0, 20),
+    "debt_sustainability_rating": (1, 5),
+    "refinancing_risk_score": (0, 100),
+    "primary_balance_pct": (-20, 15),
+    "fiscal_space_score": (0, 100),
+    "gdp_growth_pct": (-15, 20),
+    "inflation_pct": (0, 100),
+    "reserves_months_imports": (0, 30),
+    "revenue_to_gdp_pct": (1, 60),
+    "expenditure_to_gdp_pct": (2, 70),
+}
+
+
+def run_checks(df: pd.DataFrame) -> list[str]:
+    """Run plausibility checks; return list of failure messages."""
+    failures = []
+    n = len(df)
+
+    # 1. Range checks
+    for col, (lo, hi) in CHECKS.items():
+        if col not in df.columns:
+            failures.append(f"MISSING COLUMN: {col}")
+            continue
+        below = (df[col] < lo).sum()
+        above = (df[col] > hi).sum()
+        if below > 0:
+            failures.append(f"RANGE: {col} has {below} values below {lo}")
+        if above > 0:
+            failures.append(f"RANGE: {col} has {above} values above {hi}")
+
+    # 2. External + domestic ≈ 100
+    if "external_debt_pct" in df.columns and "domestic_debt_pct" in df.columns:
+        comp_sum = df["external_debt_pct"] + df["domestic_debt_pct"]
+        bad = ((comp_sum - 100).abs() > 0.1).sum()
+        if bad > 0:
+            failures.append(f"COMPOSITION: {bad} rows where external+domestic ≠ 100%")
+
+    # 3. Expenditure > revenue (fiscal deficit)
+    if "expenditure_to_gdp_pct" in df.columns and "revenue_to_gdp_pct" in df.columns:
+        exp_lt_rev = (df["expenditure_to_gdp_pct"] < df["revenue_to_gdp_pct"]).sum()
+        if exp_lt_rev / n > 0.85:
+            failures.append(
+                f"FISCAL: {exp_lt_rev}/{n} ({100*exp_lt_rev/n:.1f}%) rows have expenditure < revenue (unusual for SSA)"
+            )
+
+    # 4. No NaN/inf
+    nan_counts = df.isna().sum()
+    for col, cnt in nan_counts.items():
+        if cnt > 0:
+            failures.append(f"NaN: {col} has {cnt} missing values")
+
+    inf_cols = df.select_dtypes(include=[np.number]).columns
+    for col in inf_cols:
+        inf_cnt = np.isinf(df[col]).sum()
+        if inf_cnt > 0:
+            failures.append(f"INF: {col} has {inf_cnt} infinite values")
+
+    # 5. Row count
+    if len(df) < 25_000:
+        failures.append(f"SIZE: Dataset has only {len(df)} rows (expected ≥25K)")
+
+    # 6. All 12 countries present
+    expected_countries = {
+        "Nigeria", "Kenya", "Ghana", "South Africa", "DRC", "Ethiopia",
+        "Tanzania", "Rwanda", "Mozambique", "Zambia", "Senegal", "Cote d'Ivoire",
+    }
+    actual = set(df["country"].unique())
+    missing = expected_countries - actual
+    if missing:
+        failures.append(f"COUNTRIES: Missing: {missing}")
+
+    # 7. All 3 scenarios present
+    expected_scenarios = {"baseline", "debt_consolidation", "debt_distress"}
+    actual_scenarios = set(df["scenario"].unique())
+    missing_sc = expected_scenarios - actual_scenarios
+    if missing_sc:
+        failures.append(f"SCENARIOS: Missing: {missing_sc}")
+
+    # 8. Scenario-specific plausibility
+    for sc in ["baseline", "debt_consolidation", "debt_distress"]:
+        sub = df[df["scenario"] == sc]
+        if len(sub) == 0:
+            continue
+        mean_dsr = sub["debt_service_ratio_pct"].mean()
+        if sc == "debt_distress" and mean_dsr < 15:
+            failures.append(f"SCENARIO: {sc} mean DSR={mean_dsr:.1f}% seems too low")
+        if sc == "debt_consolidation" and mean_dsr > 30:
+            failures.append(f"SCENARIO: {sc} mean DSR={mean_dsr:.1f}% seems too high")
+
+    return failures
+
+
+def generate_diagnostic_plots(df: pd.DataFrame):
+    """Generate 8-panel diagnostic figure."""
+    os.makedirs(PLOT_DIR, exist_ok=True)
+    plt.style.use("seaborn-v0_8-whitegrid")
+    fig = plt.figure(figsize=(20, 16))
+    gs = GridSpec(3, 3, figure=fig, hspace=0.35, wspace=0.30)
+    colors = {"baseline": "#2196F3", "debt_consolidation": "#4CAF50", "debt_distress": "#F44336"}
+
+    # Panel 1: Debt-to-GDP distribution by scenario
+    ax1 = fig.add_subplot(gs[0, 0])
+    for sc, c in colors.items():
+        sub = df[df["scenario"] == sc]
+        ax1.hist(sub["debt_to_gdp_pct"], bins=50, alpha=0.5, label=sc.replace("_", " ").title(), color=c)
+    ax1.set_xlabel("Debt-to-GDP (%)")
+    ax1.set_ylabel("Count")
+    ax1.set_title("A) Debt-to-GDP Distribution by Scenario")
+    ax1.legend(fontsize=8)
+
+    # Panel 2: Debt service ratio by country
+    ax2 = fig.add_subplot(gs[0, 1])
+    country_order = df.groupby("country")["debt_service_ratio_pct"].median().sort_values(ascending=False).index
+    bp = df.boxplot(
+        column="debt_service_ratio_pct", by="country", ax=ax2,
+        vert=True, patch_artist=True, showfliers=False,
+        positions=range(len(country_order)),
+    )
+    ax2.set_xticklabels(country_order, rotation=45, ha="right", fontsize=7)
+    ax2.set_title("B) Debt Service Ratio by Country")
+    ax2.set_xlabel("")
+    plt.sca(ax2)
+    plt.title("B) Debt Service Ratio by Country")
+
+    # Panel 3: External vs Domestic composition
+    ax3 = fig.add_subplot(gs[0, 2])
+    for sc, c in colors.items():
+        sub = df[df["scenario"] == sc]
+        ax3.scatter(sub["external_debt_pct"], sub["domestic_debt_pct"],
+                    alpha=0.05, s=5, color=c, label=sc.replace("_", " ").title())
+    ax3.plot([0, 100], [100, 0], "k--", alpha=0.3, linewidth=1)
+    ax3.set_xlabel("External Debt (%)")
+    ax3.set_ylabel("Domestic Debt (%)")
+    ax3.set_title("C) External vs Domestic Debt Composition")
+    ax3.legend(fontsize=8, markerscale=10)
+
+    # Panel 4: Debt sustainability rating distribution
+    ax4 = fig.add_subplot(gs[1, 0])
+    for sc, c in colors.items():
+        sub = df[df["scenario"] == sc]
+        counts = sub["debt_sustainability_rating"].value_counts().sort_index()
+        ax4.bar(counts.index + (0.2 if sc == "debt_consolidation" else (-0.2 if sc == "baseline" else 0)),
+                counts.values, width=0.2, alpha=0.7, color=c, label=sc.replace("_", " ").title())
+    ax4.set_xlabel("Debt Sustainability Rating (1=Low Risk, 5=Distress)")
+    ax4.set_ylabel("Count")
+    ax4.set_title("D) Debt Sustainability Rating Distribution")
+    ax4.legend(fontsize=8)
+
+    # Panel 5: Concessional vs interest rate
+    ax5 = fig.add_subplot(gs[1, 1])
+    sample = df.sample(min(5000, len(df)), random_state=42)
+    for sc, c in colors.items():
+        sub = sample[sample["scenario"] == sc]
+        ax5.scatter(sub["concessional_debt_pct"], sub["average_interest_rate_pct"],
+                    alpha=0.3, s=8, color=c)
+    ax5.set_xlabel("Concessional Debt (%)")
+    ax5.set_ylabel("Average Interest Rate (%)")
+    ax5.set_title("E) Concessional Debt vs Interest Rate")
+
+    # Panel 6: Refinancing risk vs maturity
+    ax6 = fig.add_subplot(gs[1, 2])
+    for sc, c in colors.items():
+        sub = sample[sample["scenario"] == sc]
+        ax6.scatter(sub["average_maturity_years"], sub["refinancing_risk_score"],
+                    alpha=0.3, s=8, color=c, label=sc.replace("_", " ").title())
+    ax6.set_xlabel("Average Maturity (Years)")
+    ax6.set_ylabel("Refinancing Risk Score")
+    ax6.set_title("F) Refinancing Risk vs Debt Maturity")
+    ax6.legend(fontsize=8, markerscale=5)
+
+    # Panel 7: Primary balance vs fiscal space
+    ax7 = fig.add_subplot(gs[2, 0])
+    for sc, c in colors.items():
+        sub = sample[sample["scenario"] == sc]
+        ax7.scatter(sub["primary_balance_pct"], sub["fiscal_space_score"],
+                    alpha=0.3, s=8, color=c)
+    ax7.set_xlabel("Primary Balance (% GDP)")
+    ax7.set_ylabel("Fiscal Space Score")
+    ax7.set_title("G) Primary Balance vs Fiscal Space")
+    ax7.axvline(0, color="gray", linestyle="--", alpha=0.5)
+
+    # Panel 8: Scenario comparison heatmap (mean values)
+    ax8 = fig.add_subplot(gs[2, 1:])
+    metrics = [
+        "debt_to_gdp_pct", "debt_service_ratio_pct", "concessional_debt_pct",
+        "average_interest_rate_pct", "refinancing_risk_score", "fiscal_space_score",
+        "primary_balance_pct", "debt_sustainability_rating",
+    ]
+    scenario_means = df.groupby("scenario")[metrics].mean()
+    # Normalize for heatmap
+    normed = (scenario_means - scenario_means.min()) / (scenario_means.max() - scenario_means.min() + 1e-9)
+    im = ax8.imshow(normed.values, aspect="auto", cmap="RdYlGn_r", vmin=0, vmax=1)
+    ax8.set_xticks(range(len(metrics)))
+    ax8.set_xticklabels([m.replace("_", "\n") for m in metrics], fontsize=7, rotation=0, ha="center")
+    ax8.set_yticks(range(len(scenario_means.index)))
+    ax8.set_yticklabels([s.replace("_", " ").title() for s in scenario_means.index])
+    ax8.set_title("H) Scenario Comparison (Normalized Mean Values)")
+    # Add text annotations
+    for i in range(len(scenario_means.index)):
+        for j in range(len(metrics)):
+            val = scenario_means.values[i, j]
+            ax8.text(j, i, f"{val:.1f}", ha="center", va="center", fontsize=6,
+                     color="white" if normed.values[i, j] > 0.6 else "black")
+
+    fig.suptitle("African Public Debt Management — Diagnostic Plots", fontsize=16, fontweight="bold", y=0.98)
+    plot_path = os.path.join(PLOT_DIR, "diagnostic_plots.png")
+    fig.savefig(plot_path, dpi=150, bbox_inches="tight")
+    plt.close(fig)
+    print(f"Diagnostic plots saved to {plot_path}")
+
+
+def main():
+    combined_path = os.path.join(OUTPUT_DIR, "all_scenarios.csv")
+    if not os.path.exists(combined_path):
+        print(f"ERROR: {combined_path} not found. Run generate_dataset.py first.")
+        sys.exit(1)
+
+    df = pd.read_csv(combined_path)
+    print(f"Loaded {len(df)} records from {combined_path}")
+    print(f"  Countries: {df['country'].nunique()}")
+    print(f"  Scenarios: {df['scenario'].unique()}")
+    print(f"  Columns: {len(df.columns)}")
+
+    # Run plausibility checks
+    print("\n── Plausibility Checks ──")
+    failures = run_checks(df)
+    if failures:
+        for f in failures:
+            print(f"  ✗ {f}")
+        print(f"\n{len(failures)} check(s) failed.")
+    else:
+        print("  ✓ All checks passed.")
+
+    # Summary statistics
+    print("\n── Summary Statistics ──")
+    numeric_cols = df.select_dtypes(include=[np.number]).columns.tolist()
+    summary = df.groupby("scenario")[numeric_cols].agg(["mean", "std"]).round(2)
+    print(summary.to_string(max_cols=12))
+
+    # Generate plots
+    print("\n── Generating Diagnostic Plots ──")
+    generate_diagnostic_plots(df)
+
+    return 0 if not failures else 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())