A fair comparison starts with the rules.

Original official records.

Official U.S. government records, including the White House, Pentagon and federal statistical agencies, are the evidence base. For economic statistics we use BEA data, republished by the Federal Reserve Bank of St. Louis, with the original producer and source update identified. The Defense Casualty Analysis System is a Pentagon source; the U.S. Army-hosted Desert Storm article is explicitly credited to DoD News. Media reporting and expert presidential polls are not used to score these records.

An official announcement establishes what an administration said. It does not independently prove its interpretation or causal claims. We distinguish official assertions, published statistical observations and our calculations.

Same measure. Stated window.

GDP compares 12 completed four-year calendar windows from 1977 through 2024. Trump II is incomplete and shown separately. The source is BEA annual real GDP levels in billions of chained 2017 dollars, via the Federal Reserve’s GDPCA series, updated April 9, 2026. This is a frozen edition, not a claim to incorporate subsequent revisions. Each term includes two prior years and the following year for context; those outside years do not enter its rank. Border comparisons use July apprehensions in four consecutive years. Crime displays the White House’s reported 2025 annual changes; a historical presidential ranking is not yet supported.

For the default window, the baseline is the annual GDP level in the year before the term and the endpoint is year 4. Annualized growth = [(end ÷ baseline)^(1 ÷ years) − 1] × 100; total change = (end ÷ baseline − 1) × 100. These are calculated from levels, not by averaging or compounding rounded growth percentages. The optional years 2–4 view uses year 1 as its baseline and a three-year exponent. It is a sensitivity check, not a causal adjustment or a rule awarding year 1 to the predecessor. Neither view isolates policy effects. The Federal Reserve describes long and variable monetary-policy lags; that is not a universal lag for all presidential policies.

Ranks describe only the listed observations. GDP ranks use displayed one-decimal rates. Ties share a rank and skip subsequent positions. Percentage change is (later − earlier) ÷ |earlier| × 100; zero or missing baselines produce no result. Outcomes reflect inherited conditions, external events, Congress and many actors beyond the president.

Define completion before counting it.

Each included pledge needs an official record identifying the campaign commitment, a specific completion criterion and dated outcome evidence. “Completed” meets the criterion. “Partial / limited” delivers a narrower or incomplete result. “Not fulfilled as defined” fails that criterion at the stated checkpoint. “Not verified” means evidence is insufficient, not that the pledge failed.

The sample rate is completed ÷ assessed pledges × 100. Partial and unfulfilled pledges stay in the denominator. Unverified entries are excluded and separately counted. The current seven-item collection is unequal and nonrepresentative; its percentages must not be treated as total campaign fulfillment rates or used to rank presidents.

Scope every casualty.

Duration uses the stated operation boundaries, not an assumed definition of the entire war. Hostile deaths include killed in action and other hostile categories; killed in action is a subset, not an additional count. Total deaths add hostile and non-hostile deaths. Wounded in action is separate. DCAS totals are as of September 23, 2026 and may be revised.

Operations spanning multiple administrations remain attributed to all presidents in that window. OEF counts include places outside Afghanistan. No missing figure is converted to zero. Official opposing-force estimates remain estimates, and civilian harm is reported separately.

An army-size description is not a global composite power ranking. We require a source referring to the forces at the time of the conflict, with the measure identified.