Meta-estimates of dead and wounded in the Russia-Ukraine war (2022-)

Live estimates

As of 2026-03-20:

• Estimated deaths: 236,000 to 467,000 (or, roughly 332,000)
• Estimated casualties: 1,089,000 to 1,475,000 (or, roughly 1,267,000)

Our estimate combines public and leaked intelligence on military casualties in the Russia–Ukraine war with geospatial data on conflict intensity derived from our war-fire system and calculations of shifts in areas of control. We fit a generalized additive model (using R’s mgcv package) with three types of predictors:

• Temporal trend: days since 24 February 2022
• Changes in territory: cumulative net changes in area controlled by each side
• War intensity: a tensor-product smooth of cumulative war-fire activity (the sum of daily logged firing events) and cloud cover, both stratified by area of control

From this GAM we derive daily point estimates, 95% confidence intervals and prediction intervals for casualties and fatalities dating back to the start of the invasion. At present we report only Russian losses, as there are too few credible data on Ukrainian fallen to support a robust meta-estimate.

Limitations

We do not have ground-truth data on Russian (or Ukrainian) losses in Ukraine. Instead, we provide an estimate-of-estimates, which should be interpreted with care. Our model infers what the total dead, and dead and wounded, might be if the underlying estimates are correct—but we cannot guarantee that they are. The precise timing of losses should be treated with special care.

Prerequisites

• R ≥ 4.0
• R packages: tidyverse, lubridate, mgcv, scales, zoo

Data Layout

scripts/
└─ meta\_estimate.R

source-data/
├─ Soldier\_deaths\_casualties\_estimates.csv
├─ UK\_MoD\_monthly\_estimates.csv
├─ meduza\_2024.csv
├─ meduza\_2025\_weekly.csv
├─ strikes\_by\_location\_and\_day.csv
└─ area\_assessed\_as\_controlled.csv

output-data/
├─ meta-estimate-casualties.csv
└─ meta-estimate-deaths.csv

plots/
└─ meta-estimate.png

Usage

Rscript scripts/meta_estimate.R

• Halts if no data newer than 3 months is found
• Generates CSVs in output-data/ and plot in plots/

Workflow

1. Load & clean multiple source CSVs
2. Merge estimates (deaths vs. casualties)
3. Load covariates (fires, cloud cover, territory changes)
4. Fit GAM with splines & tensor terms
5. Predict through today, calculate 95 % CI & PI
6. Export CSVs & save plot

Outputs

• output-data/meta-estimate-casualties.csv
• output-data/meta-estimate-deaths.csv
• plots/meta-estimate.png

Licence

This software is published by The Economist under the MIT licence. The data generated by The Economist are available under the Creative Commons Attribution 4.0 International License.

The data and files that we have generated from open sources are freely available for public use, as long as The Economist is cited as a source.

Suggested citation

The Economist and Solstad, Sondre (corresponding author), 2025. The Economist's meta-estimate of casualties in the Russia-Ukraine war. First published in the article "Russia's summer offensive looks like its deadliest yet", The Economist, July 9th, 2025.