Abstract

This page documents the implemented calculator, not an aspirational model. It describes the public APIs, static datasets, mathematical transformations, fallback rules, uncertainty treatment, and audit notes that currently drive the application.

The system treats war-cost estimation as a transparent accounting exercise. Every user-facing number must either come from a named public source or from a deterministic transformation applied to those sources. When live APIs fail or provide no coverage, the code falls back to explicit local datasets rather than silently imputing values.

The calculator is intentionally conservative in scope. It prices military operations, armaments and intercepts, economic dislocation, humanitarian displacement support, direct casualty cost, and reconstruction. It does not claim to price the full social cost of war, and it explicitly excludes nuclear escalation, long-run trauma, ecological damage, alliance cascades, and other second-order effects that are either methodologically unstable or not yet parameterized in code.

Data Pipeline

The codebase is organized as a small Next.js application: static pages at the top, server routes in the middle, and deterministic calculation modules beneath them.

Internal Routes

Country Assembly Pipeline

The calculation route starts by building two enriched country objects. Geographic metadata comes from REST Countries, live indicators come from World Bank, GDP can fall back to IMF DataMapper, military and sanctions context can fall back to local datasets, and missing data-sparse states are explicitly labeled through the hasStaticFallback flag.

Added Calculations

Recent live-data refinements do not replace the base model. They sit on top of the prior architecture to improve trade realism, inflation handling, commodity scaling, and target-country fragility sensitivity.

Live APIs

The calculator mixes live public APIs with versioned local datasets. Live APIs are used for recency; local files are used for coverage, calibration, and transparent fallback behavior.

API Endpoint Map

These are the actual upstream endpoint patterns the code calls today. The app wraps them with cache control, graceful fallback, and input normalization before they influence any result.

Static Datasets

Versioned local datasets provide coverage, calibration anchors, and transparent fallback behavior when live APIs are unavailable or return null.

Military Model

The military module is anchored to direct operational spending benchmarks from Watson Institute case studies and then scaled by aggressor budget, scenario class, distance, and attrition. The Watson anchor for a conventional war is ~$200M/day ≈ cost of 50 Tomahawk cruise missiles daily. Air-campaign munitions, aircraft packages, and defensive intercepts are handled in the armaments module, keeping this module focused on operations, personnel, logistics, and wear. It does not attempt to reproduce veterans care, interest on war debt, or homeland-security spillovers.

Notes: Operational totals are decomposed into personnel (35%), operations and logistics (40%), munitions (20%), and C3ISR (5%), with scenario-specific equipment attrition added on top.

Economic Model

The economic module combines bilateral trade disruption, target-country GDP contraction, capital flight, sanctions drag on the aggressor when justified by literature, and a commodity shock layer for globally important producers.

Notes: Live Comtrade trade volume overrides the static pair table when available. Oil, gas, and wheat shocks can scale with live FRED prices. ACLED can add a modest target fragility overlay to GDP loss and capital flight.

Humanitarian Model

The humanitarian module estimates displacement support, emergency healthcare, and direct casualty cost. It applies country-specific or regional UNHCR ratios for most scenarios, uses observed population-share calibration for sustained air campaigns, and prices casualties through a conservative GDP-per-capita human-capital method. The per-person-year support cost of $1,500 ≈ UNHCR average emergency response cost per displaced person covers humanitarian assistance and emergency healthcare.

Notes: The module splits displaced people into IDPs and cross-border refugees according to the observed UNHCR ratio mix. ACLED can modestly scale displacement when recent target-country violence is already elevated. Direct casualty cost uses GDP per capita x 100 as a human-capital proxy, with injury cost set to 15% of VSL.

Reconstruction Model

Reconstruction is modeled as a sublinear function of GDP so that richer countries do not mechanically receive implausible rebuild bills just because their output base is larger. A 30% overlap discount is applied because part of the destruction is already expressed as lost GDP in the economic module.

Notes: The 20B reference point corresponds to the Afghanistan calibration point used in the code. Opportunity-cost widgets are derived from reconstruction totals but are displayed outside the main war total.

Armaments Model

The armaments module prices weapons procurement, munitions consumption, equipment attrition, and defensive intercept costs — four buckets that were absent from the original model. It uses live military expenditure from the World Bank API with SIPRI Milex as offline fallback, NATO equipment-percentage data to derive a procurement fraction, a static unit-cost table sourced from DoD annual budget documents, and scenario-based force-package tables calibrated to real conflicts.

Notes: budgetScalar uses a power-law exponent of 0.75 (diminishing returns — larger budgets buy more but not linearly). Equipment fraction defaults to 20% of military spend for non-NATO states; NATO Table 8a values used when available. Unit costs cover 22 weapon categories from cruise missiles ($2M ≈ annual salary of 40 US teachers) to aircraft carriers ($13B ≈ GDP of Iceland). Intercept costs are anchored to CSIS Iran 2026 reporting: $1.7B in the first 100 hours, then checked against later reporting of roughly 700 ballistic missiles and 3,600 drones in the early campaign ($395K average per threat ≈ 5× cost of an Iron Dome Tamir intercept). Data sources: SIPRI Milex 2024 (135 countries), NATO Defence Expenditure 2025, Bruegel US Foreign Military Sales 2008–2025, DoD Program Acquisition Costs FY2024, GAO-24-106649 Ukraine Weapon Replacement Study.

Revenue Counterfactual

Revenue is not subtracted from the headline projected cost. It is shown as a deliberately optimistic counterfactual: what an aggressor might hope to extract if it wins, holds territory, and manages to keep production online despite sabotage, sanctions, and infrastructure damage.

Notes: Capture rates are scenario-specific: 0% for skirmish, 15% for conventional war, and 50% for occupation. The code treats this section as best-case and low-confidence by design.

Aggregation and Uncertainty

The app does not run a Monte Carlo engine. Instead, each module defines its own conservative range. Headline cost and economic impact are aggregated separately, which keeps the accounting explicit and avoids folding macroeconomic spillovers into the top-line war bill.

Scenario durations are normalized archetypes rather than event-specific backtests: 0.15 years for air_campaign, 0.2 years for skirmish, 1.5 years for conventional war, and 10 years for occupation at point estimate. Armaments ranges are wider than other modules due to force-package composition uncertainty.

Calibration — Operation Epic Fury (Iran, 2026)

On February 28, 2026, the United States and Israel launched a sustained air campaign against Iran. The first validation pass used Day 17 reporting; the late-April reassessment can now compare the model against a 38–39 day campaign and a ceasefire that took effect on April 8. The 17-day cost reached $14–16.5B ≈ what the US spends on Head Start for 3 years.

Conflict Profile

Key Discovery: Cost Is Two-Sided

The CSIS breakdown of the first 100 hours revealed that defensive intercept costs ($1.7B ≈ 4 Nimitz-class refueling overhauls, 46%) exceeded offensive strike munitions ($1.5B, 40%) in the opening phase. The original model only priced what the aggressor spends attacking. The cost of neutralizing the counter-attack was entirely absent.

Five Gaps Found — Five Fixes Applied

Pre / Post Results vs Late-April Data

Lesson: the original 18-day strike comparison still matched the early direct-cost snapshot, but late-April data makes the accounting boundary visible. Public US cost estimates are narrow federal military outlays; this calculator’s headline also includes humanitarian and reconstruction costs. The human-toll and economic-impact modules now match the updated evidence better than the direct military spending module, which remains high for the observed 39-day campaign.

Calibration Sources

Audit Notes

This repository was checked against the shipped data files and the running model logic. The goal was not to prove truth, but to document what the current code can defensibly claim.

1. Import-time JSON validation is enforced in validated.ts for bilateral trade pairs and commodity producer datasets, including schema shape, ISO alpha-3 keys, numeric bounds, and duplicate-pair normalization.
2. The repository currently ships 75 bilateral trade pairs, 26 country displacement cases, 8 regional displacement defaults, 84 SIPRI entries, 14 static fallback countries, 8 sanctions regimes, and 32 commodity shock records.
3. World Bank retrieval now requests the 10 most recent observations so the app can use the latest non-null value instead of defaulting to a fallback too early.
4. The Iran calibration table uses reparameterized 39-day and 55-day comparisons. Other historical claims remain archetypal scenario checks rather than event-level backtests.
5. When FRED is unavailable, the model remains runnable. Commodity shocks stay on 2023 baselines and the CPI scalar remains 1.0, which preserves deterministic behavior instead of fabricating live prices.

Scope Limits

Exclusions are a methodological choice rather than an omission. The model only prices what the code can currently source, parameterize, and explain line by line.

• Nuclear escalation, WMD use, and strategic deterrence failure.
• Cyber operations, satellite warfare, or electromagnetic-spectrum disruption costs.
• Population-scale trauma, long-run disability, and demographic scarring.
• Alliance activation, extended coalition war finance, and treaty spillovers.
• Environmental destruction, landmine clearance, and remediation of contaminated land or water.
• Brain drain, educational loss, and long-run human-capital depreciation.
• Shadow-economy disruption and illicit finance networks in already-sanctioned states.

Open Calculator → Start with the calculator for the result, then inspect each category line item, then return here for the underlying model assumptions and data pipeline.