EPA FuelEconomy.gov Web Services API Data

Every fuel-economy rating the US government publishes has an addressable version behind it: a structured feed served from fueleconomy.gov itself, administered by Oak Ridge National Laboratory for the US Department of Energy and EPA. The EPA FuelEconomy.gov Web Services API is that feed - drill-down menus that walk from model year to make to model to option combination, then a full specification record for the exact vehicle you land on.

Three record families ride on it. Vehicle specifications carry the numbers consumers see on window stickers - city, highway and combined MPG, CO2 per mile, displacement, cylinders, transmission, drive type, vehicle class - plus the electrified fields modern lineups demand. Emissions records attach per-vehicle SmartWay and greenhouse-gas scores against named standards. And a separate family exposes driver-submitted real-world MPG from the My MPG program, the crowd-sourced counterweight to laboratory ratings.

For automobile manufacturers, that combination is unusual: official ratings you can cite and owner-reported consumption you can contrast them with, resolvable down to a single option code.

A menu walk narrows the universe before any heavy payload arrives - year, then make, then model, then options. The destination is one vehicle record:

id          : 31873
year        : 2023
make        : Toyota
model       : Camry
VClass      : Midsize Cars
drive       : Front-Wheel Drive
trany       : Automatic 8-spd
displ       : 2.5     cylinders : 4
fuelType1   : Regular

cty         : 28      ucty : 33
hwy         : 39      uhwy : 47
comb        : 32      ucmb : 38
co2         : 277  (grams/mile)

An electrified configuration swaps the second block for its own vocabulary:

id          : 41385
make/model  : Hyundai / Ioniq 5
fuelType1   : Electricity
charge240   : 6.3   (hours on 240V)
phevComb    : --    (combined MPGe for PHEVs)
combinedUF  : --    (utility factor, PHEVs)

And the third family looks nothing like either - it is owners reporting what their cars actually drink:

userVehicle : shared summary for one year/make/model
avgMpg      : 31.4   (driver-reported)
records     : 87     distinct contributors

Values above illustrate the shape of each record family rather than quote a live pull; your sample arrives with real vehicles scoped to the years and makes you name.

Twenty-three documented fields verified against the publisher's own web-services documentation - the dictionary below covers the core schema for vehicle, emissions and fuel-price responses. Nothing is inferred.

Geography - the United States market exclusively: every vehicle record reflects EPA ratings of US-market configurations, and emissions records are keyed to sales areas within the country. National fuel prices are likewise US-wide averages.

Temporal - model years 1984 through the current model year in the vehicle tables, the same universe as the published bulk files. User-submitted MPG records span the life of the My MPG program instead, which starts later than the ratings themselves. Annual refreshes fold each new model year into the menus once EPA certifies it.

Granularity - one specification record per vehicle configuration: a unique year, make, model and engine/transmission option. Menus enforce the hierarchy, so resolution always terminates at exactly one configuration - the property that lets a lookup product answer "this car, these numbers" without ambiguity.

API, files, or your warehouse. Daily, weekly, or hourly.

• Vehicle research and shopping tools - resolve a user's year/make/model pick into certified MPG, CO2 and class on the fly, the canonical backend for configurators and comparison widgets.
• Claim verification - check an advertiser's or competitor's quoted MPG, CO2 or SmartWay score against the government's own per-vehicle record before it ships in copy or a deck.
• Real-world versus rated gap analysis - join driver-submitted MPG records to official ratings and quantify where the label overpromises, segment by segment.
• Fleet and total-cost-of-ownership tools - per-configuration efficiency plus national fuel prices turn "what will this van cost to run" into arithmetic rather than guesswork.
• Electrification tracking - charge times, combined MPGe and utility factors give plug-in lineups their own measurable trail across recent model years.
• Machine-learning features and benchmarks - schema-stable specification records make clean training rows for efficiency-prediction models at point-lookup scale.

Developers and data-product builders get keyless JSON/XML lookups shaped for integration - the rare official feed that behaves like a product API. Competitive-intelligence and product teams get rivals' specs and emissions scores on demand for comparison content. Market researchers and consultants get verification-grade figures for client deliverables. Journalists, academics and students get the citable endpoint behind every MPG claim in consumer reporting. Data scientists get point-lookup depth here and should pair it with the sibling bulk table when modeling demands all years at once.

Four things worth knowing upfront. First, this is a point-lookup catalog: individual responses are single objects or small lists, sized for resolving one vehicle at a time. If your workload needs every configuration since 1984 in one flat table, say so on the sample request - the sibling EPA Fuel Economy Dataset (1984-2026) is the right shape and we deliver both.

Second, adjusted and unadjusted MPG are different series. The headline columns carry window-sticker-style figures consumers recognize; the u-prefixed variants carry raw dynamometer results before adjustment. Keep them apart in any pipeline.

Third, the electrified vocabulary only appears where electrification does - charge240, phevComb and utility factors are populated on EV and PHEV configurations and null elsewhere, and dual-fuel vehicles fill both fuel-type fields. Treat empty electrified fields as structure, not missingness.

Fourth, the publisher documents no formal usage policy for heavy programmatic use, so sustained high-volume pulls should be raised with us at sample time - we will scope the delivery pattern (batch files or warehouse sync) that keeps your load off the live service entirely.