Data API Access

Programmatic access to PACE forecast data via stable, scrapable URLs. Updated automatically on the 1st of each month.

These URLs remain constant - only the data updates monthly.

Quick Start

All datasets are available as direct downloads via stable URLs. Use these URLs in your scripts, notebooks, or automated pipelines.

Example: Fetch 12-month forecasts

# Python
import pandas as pd
df = pd.read_csv('https://raw.githubusercontent.com/conflictlab/Pace-map-risk/main/forecasts_h12.csv')

# R
df <- read.csv('https://raw.githubusercontent.com/conflictlab/Pace-map-risk/main/forecasts_h12.csv')

# curl
curl -O https://raw.githubusercontent.com/conflictlab/Pace-map-risk/main/forecasts_h12.csv

# wget
wget https://raw.githubusercontent.com/conflictlab/Pace-map-risk/main/forecasts_h12.csv

📅 Update Schedule

Forecast Generation:1st of each month at 01:00 UTC

Data Available:Within 2 hours of generation (typically by 03:00 UTC)

URL Stability:URLs never change, only content updates

Next Update:April 1, 2026 (automated)

12-Month Forecasts

Forecasts 12 months ahead (mean, min, max)

forecasts_h12.csv

Mean 12-month forecast

Download

forecasts_h12_min.csv

Lower bound (minimum scenario)

Download

forecasts_h12_max.csv

Upper bound (maximum scenario)

Download

6-Month Forecasts

Forecasts 6 months ahead (mean, min, max)

forecasts_h6.csv

Mean 6-month forecast

Download

forecasts_h6_min.csv

Lower bound (minimum scenario)

Download

forecasts_h6_max.csv

Upper bound (maximum scenario)

Download

Historical Data

Complete historical time series

Hist.csv

Full historical time series (all available UCDP data)

Download

Metadata

Information about the forecast run

forecast_metadata.json

Run date, data ranges, configuration

Download

Metadata Format

The forecast_metadata.json file contains:

{
  "run_date": "2026-03-01T01:15:43.123456",
  "data_end_date": "2026-02",
  "forecast_start_date": "2026-03",
  "h6_end_date": "2026-08",
  "h12_end_date": "2027-02",
  "training_window_months": 24,
  "historical_start_date": "1989-01",
  "historical_end_date": "2026-02",
  "total_historical_months": 445
}

Data Format

📅 Understanding Forecast Periods

Check forecast_metadata.json to see exactly what months are being forecast:

data_end_date: Last month of historical data used (e.g., "2026-02")
forecast_start_date: First month being forecast (e.g., "2026-03")
h6_end_date / h12_end_date: Final forecast month

Forecast CSVs (forecasts_h6.csv, forecasts_h12.csv):

Rows: Numbered 0 through 5 (h=6) or 0 through 11 (h=12)
Row 0 = First forecast month (see forecast_start_date in metadata)
Row 1 = Second forecast month, etc.
Columns: Countries (e.g., "Afghanistan", "Algeria", etc.)
Values: Predicted fatalities for that month

Example: If forecast_start_date is "2026-03", then row 0 = March 2026, row 1 = April 2026, etc.

Historical CSV (Hist.csv):

Rows: All available months from UCDP data (1989-01-31 onwards)
First column: Date index (YYYY-MM-DD format)
Other columns: Countries
Values: Observed fatalities for that month
Coverage: From 1989 through data_end_date (see metadata)

How to Cite

If you use these data or figures, please cite:

Schincariol, T., Frank, H., & Chadefaux, T. (2025). Accounting for variability in conflict dynamics: A pattern-based predictive model. Journal of Peace Research. https://doi.org/10.1177/00223433251330790

See full citation guidelines for additional details.

Usage Terms

License: CC BY-NC 4.0 (Attribution-NonCommercial)

Commercial Use: For commercial use, please contact us.

Rate Limiting: Please cache responses and avoid excessive requests. Data only updates monthly - fetching more frequently is unnecessary.

Support & Questions

For technical questions, data issues, or feature requests: