Methodology

01 Government Data Sources

Every input to True Driving Cost's calculations is U.S. government data (EPA, DOE, EIA, BLS and BEA); no third-party or industry data feeds are used. The site also publishes results it computes itself from that data: vehicle rankings, class comparisons, the historical price replay and simulation, and the fuel price exposure table. Each is described in this document and labelled as the site's own computation, never attributed to a government agency. The underlying data comes from four groups of sources: the FuelEconomy.gov service, the price histories and reference tables described in Visualizations, the Bureau of Labor Statistics and Bureau of Economic Analysis figures behind the ownership-costs panel (see Ownership Costs Beyond Fuel), and the FHWA state mileage data behind the mileage-comparison panel (see How Your Mileage Compares).

FuelEconomy.gov REST API, maintained jointly by the U.S. Department of Energy (DOE), Office of Energy Efficiency & Renewable Energy, and the U.S. Environmental Protection Agency (EPA). This is the primary, real-time data source for every vehicle-specific figure on the page.

• Vehicle data: fueleconomy.gov/feg/ws/index.shtml
• Fuel prices: fueleconomy.gov/ws/rest/fuelprices
• EPA fuel economy program: epa.gov/fuel-economy
• EPA Green Vehicles: epa.gov/greenvehicles

All API requests for the selected vehicle's own specs are made client-side in real time. No individual vehicle data is cached, stored, or modified by True Driving Cost.

Price histories (DOE Alternative Fuels Data Center gasoline series, EIA diesel and EIA electricity series, and the CPI-U index from EIA Monthly Energy Review Table 1.06 used to express prices in constant dollars, listed in the table under Visualizations) are not fetched live. They are written into the page from files in this project's data folder by a script (audit/price_history_data.py), and a check fails if the page and the files disagree. Each series needs one new year appended annually; see Data Freshness.

EPA/DOE Fuel Economy Guide vehicle database (Data/vehicles.csv, model year 2026) is a static snapshot, separate from the live per-vehicle service above. It is used only to build the reference population this site compares your selected vehicle against: the class distributions, the rankings, the EPA score percentiles and the tailpipe emissions quartiles. The tables are written into the page by audit/calc_ref_data.py; population counts and reference prices are listed under Visualizations.

02 API Fields Used

The following fields are retrieved from the FuelEconomy.gov /vehicle/{id} endpoint for each selected vehicle:

Fuel prices are retrieved from the /fuelprices endpoint:

03 Fuel Cost Calculation

Conventional Vehicles (Gasoline, Diesel, E85, CNG)

Fuel cost is computed using EPA city and highway fuel economy ratings, weighted by the user's specified city/highway driving mix using the EPA harmonic mean formula:

The harmonic mean is used (rather than arithmetic mean) because fuel economy in MPG is a rate (miles per gallon). The harmonic mean correctly weights the proportions of city and highway driving. This is the same formula used by the EPA to calculate combined fuel economy on vehicle window stickers.

04 Electric & Plug-in Hybrid Vehicles

Battery Electric Vehicles (BEVs)

For pure electric vehicles (where atvType = "EV" and no plug-in hybrid indicator), fuel cost is calculated using the EPA electricity consumption rating, blended by the same city/highway driving mix used for gas vehicles - EPA rates city and highway electricity consumption separately for EVs, and city driving is usually somewhat more efficient due to regenerative braking:

Where cityE/highwayE are the EPA-rated electricity consumption in kWh per 100 miles. Unlike the harmonic mean used for gas MPG, this is a straight weighted average, since kWh/100mi is already energy-per-distance rather than a rate like miles-per-gallon. The default electricity rate is sourced from the FuelEconomy.gov fuel prices API. Users can override this with their local rate.

Plug-in Hybrid Electric Vehicles (PHEVs)

For PHEVs, annual mileage is split between electric (charge-depleting) and gasoline (charge-sustaining) operation using the EPA's own utility factors, the estimated share of miles driven in charge-depleting mode, published separately for city and highway driving (cityUF, highwayUF) and blended with your city/highway mix:

The second term inside the parentheses covers plug-in hybrids that still burn some gasoline while running mostly on the battery (EPA's cityCD and highwayCD, gallons per 100 miles, blended the same way; zero for vehicles that do not). If a vehicle has no utility factors, the calculator falls back to its electric range divided by its total range, and if that is also unavailable, to gasoline-only operation using the conventional MPG ratings. The cost-per-mile rankings use their own fixed reference assumptions for PHEVs (described in their sections), not this cost model.

05 Fuel Price Matching

Each vehicle's fuelType field determines which fuel price is applied. The matching logic:

Fuel prices are fetched once per session from the FuelEconomy.gov service, which returns one national average per fuel with no regional breakdown. If the service is unavailable, the page uses fallback prices (listed under Visualizations) and says so in the estimate's caption. You can override the fuel price with your own.

State-level regional adjustment. When you select a state, the live national price is multiplied by that state's price divided by the national price in the EIA weekly regular-gasoline series (week and definitions listed under Visualizations). Electricity prices are not adjusted by state.

06 User Parameters & Validation

Users provide three parameters that shape the analysis:

Additional override parameters (accessible via the results interface):

All inputs are validated with a clamp(value, min, max) function at every entry point (dropdown selection, popover input, and programmatic update) to prevent invalid calculations.

07 Visualizations

All charts are drawn in your browser on HTML5 canvas; no external charting library is used. Figures are numbered as on the calculator page.

Fig. 1: Cost of ownership by year

One stacked bar per year of your ownership period. Fuel is calculated for your vehicle at today's price held flat. Depreciation is the loss in value in that year: BLS depreciation rates by vehicle age, applied to the average purchase price (or your own price) and compounded, so it differs from year to year. Other typical costs are the four BLS per-vehicle averages described in Ownership Costs Beyond Fuel, constant per year. A table below the chart lists the same figures by year with cumulative totals and the simulated 25th to 75th percentile of cumulative fuel cost.

Fig. 2: Driving mix sensitivity

Total fuel cost at every city/highway mix from 0% city to 100% city, recalculated at 2% steps. Your mix is marked, and the lowest and highest costs are labelled.

Fig. 2b, effective fuel economy by mix. Directly below Fig. 2, this vehicle's EPA city and highway ratings (MPG or MPGe) and the effective rating at every mix between them, on the same 0% to 100% city axis: the mechanism behind Fig. 2's cost curve. Your mix is marked; the city and highway endpoints are omitted from the label when your mix is exactly 0% or 100%, since "Your mix" would then restate the same number.

Fig. 3: Fuel price scenarios (historical replay)

Every run of consecutive annual price changes in the price history of your vehicle's fuel, of length (ownership years − 1), is applied to today's price after a first year at today's price. The chart shows the run with the lowest total cost, the run with the median total cost, the run with the highest total cost, and the flat-price path. Battery-electric vehicles use the residential electricity history, diesel vehicles the diesel history, plug-in hybrids the gasoline and electricity histories together (each run uses the same calendar years for both), and other vehicles the regular gasoline history. These are past sequences applied to today's price, not forecasts. A one-year period has no later year, so only the flat path is shown.

Fig. 4: Sensitivity

Each of four inputs (annual miles, city/highway mix, fuel or electricity price, and vehicle efficiency) is changed by ±20% (or ±20 percentage points for mix, limited to 0–100%) while the others stay at your settings; the resulting low and high total costs are drawn as bars ranked by swing. Efficiency is changed on the underlying city and highway rating (or kWh per 100 miles for electric vehicles), not by scaling cost directly, since cost is inversely related to fuel economy.

Fig. 5: Efficiency gauge

City, highway and effective combined MPG (MPGe for electric vehicles) at your mix, drawn on a track from 75% of the lower rating to 125% of the higher one. The caption states whichever of city or highway is higher for the vehicle. Next to it, the combined rating is placed against its EPA body-style class in the model year 2026 reference population.

Fig. 6: Cost against CO₂ by vehicle class

The median cost per mile against the median tailpipe CO₂ per mile of each of 12 EPA body-style classes (bubble size is the number of vehicles), from the model year 2026 reference population, with a fitted line and the Pearson correlation of the 12 class medians. Your vehicle is plotted at its own cost and emissions. Not shown for battery-electric vehicles.

Fig. 7: Position among model year 2026 vehicles

Box plots of cost per mile by class (boxes 25th to 75th percentile, whiskers 5th to 95th, tick at the median) with your vehicle marked, and up to six exact-rank cards: cost per mile in class, overall and among vehicles with the same powertrain (battery electric, plug-in hybrid, hybrid, diesel or gasoline); MPG or MPGe among the same powertrain; MPG in class; and tailpipe CO₂ per mile in class (rank 1 is the cheapest, most efficient or cleanest). A rank is 1 plus the number of reference vehicles that are strictly better, so ties share a rank. Cost per mile is the reference price divided by combined MPG or MPGe (33.7 kWh per gallon-equivalent); plug-in hybrids blend gasoline and electric operation 50/50 because the Fuel Economy Guide does not publish a per-vehicle electric share. This compares your vehicle against the current model year 2026 lineup, not vehicles of its own model year. MPG and CO₂ are ranked within a class or powertrain only, because comparing a truck against a compact car on those measures is not like for like.

Fig. 8: EPA comparison

Annual fuel cost on three bases: EPA's estimate for this vehicle (fuelCost08, 15,000 miles a year, 55% city), your own parameters, and an average new vehicle on EPA's basis (EPA's published five-year youSaveSpend divided by 5, added to fuelCost08). The three figures use different assumptions and are not three measurements of one quantity.

Fig. 9: Price uncertainty (Monte Carlo simulation)

1,500 simulated cost paths. Year 1 is priced at today's price; in each later year the annual price change is taken from a randomly chosen run of 3 consecutive recorded years of the price history of your vehicle's fuel (the same series as Fig. 3; a moving-block bootstrap, chosen because year-to-year price changes are persistent, see the assumption check below), and the fuel cost for that year is recalculated. The random generator is seeded from your vehicle and settings, so the same inputs always give the same result. The chart shows the 10th, 25th, 50th, 75th and 90th percentiles of cumulative cost at each year. The flat-price estimate differs from the simulated median because the flat estimate assumes no change after year 1 while the simulated paths apply changes whose average is positive; the page states the mean and median change of the series it used. The range describes outcomes if annual price moves resemble the history; it is not a forecast and not a confidence interval.

Assumption check: are annual price changes independent? Independent one-year draws would treat each year's change as unrelated to the earlier years' changes. The autocorrelation of the recorded changes (the correlation between a year's change and the change a given number of years earlier) shows that this is not a safe assumption for every series. Electricity is clearly persistent: its autocorrelation is positive and well outside the 95% band at every lag shown. For gasoline and diesel the coefficients are smaller and sit at the edge of the band, with slightly negative values at lag 2. The simulation therefore takes runs of 3 consecutive recorded years, which preserves persistence within a 3-year window where it exists without imposing it where it is weak.

Sensitivity to the draw method. Average annual fuel cost over the ownership period as a multiple of holding today's price flat (1.00 = flat), from 20,000 simulated paths per row. Between independent draws and the 3-year blocks used, the 10th percentile differs by at most 0.04, the median by at most 0.01 and the 90th percentile by at most 0.09 across the rows below. Block draws widen the upper end for electricity and gasoline; for diesel, which has only 30 annual changes, the 10-year upper end is lower under blocks. Five-year blocks are shown only to check that the choice of 3 is not driving the result.

Source: annual price series in audit/price_history_data.json (built from the files listed under Price histories); script audit/mc_robustness.py (fixed random seed). Simulated cells are Monte Carlo estimates and can differ by about 0.01 to 0.02 between runs of different software versions.

Fig. 9b, distribution. A 30-bin histogram of the 1,500 simulated total costs with a Gaussian kernel density estimate (Silverman bandwidth, h = 0.9·min(σ, IQR/1.34)·n^−1/5) and a normal curve of the same mean and variance. The flat-price estimate is marked, and the page states its percentile among the simulated outcomes. Fig. 9c, cumulative distribution. The empirical cumulative distribution of the same 1,500 totals with P10, median and P90 marked. Fig. 9d, price trend. An ordinary-least-squares fit of ln(price) on year to the nominal history of your vehicle's fuel with a 95% prediction interval (SE = s·√(1 + 1/n + (x−x̄)²/S_xx)), extended through your ownership window as an extrapolation of the fitted line and applied to your vehicle's consumption; it is not a forecast. Fig. 9e, tail risk. Value at risk at the 95th and 99th percentiles and conditional value at risk (the mean of the simulated totals at or above the 95th percentile), computed from the same 1,500 totals.

Reference data, prices and price histories

Reference population. Model year 2026 rows of the EPA/DOE Fuel Economy Guide database (Data/vehicles.csv), leaving out fuel-cell vehicles and rows outside the 12 body-style classes: 1,162 vehicles with a cost per mile, 841 non-electric vehicles with a tailpipe CO₂ figure and 841 with a combined MPG. The EPA score percentiles use all 1,172 model year 2026 rows.

Reference prices for cost per mile in the rankings are the 2025 annual U.S. averages: regular, midgrade and premium from the EIA weekly series, diesel from the EIA annual series, electricity from EIA Monthly Energy Review Table 1.6.

Price histories used by the historical replay, the simulation and the exposure table. All are nominal annual averages; the statistics are computed on year-over-year changes.

Fallback fuel prices apply only when live prices cannot be loaded: regular $3.10, midgrade $3.67, premium $4.03, diesel $3.66 and electricity $0.173 (2025 EIA annual averages); E85 $3.31, CNG $3.08 per gasoline-gallon equivalent and propane $3.59 (April 2026 national averages, DOE Clean Cities and Communities Alternative Fuel Price Report).

State fuel prices are the latest weekly values in the EIA regular-gasoline series (week of 2026-10-12, U.S. average $4.331): a state's own series where EIA publishes one (California, Colorado, Florida, Massachusetts, Minnesota, New York, Ohio, Texas, Washington), otherwise its PADD region or sub-region.

08 EPA Scores & Ratings

Three EPA ratings are displayed directly from the API with no modification:

• EPA Fuel Economy Score (feScore): 1 to 10 scale relative to all vehicles. Higher is better. Rendered as an animated score ring.
• EPA Greenhouse Gas Rating (ghgScore): 1 to 10 scale based on CO₂ tailpipe emissions. Higher is better. Rendered as an animated score ring.
• CO₂ Tailpipe Emissions (co2TailpipeGpm): EPA-measured grams per mile, shown on a spectrum from 0 g/mi to 750 g/mi and classified against the quartiles of the model year 2026 reference population (counts under Visualizations). This classification is the site's own computation from the Fuel Economy Guide database, not an EPA-published average.

Both scores are also shown as the percentage of model year 2026 vehicles scoring strictly lower, computed from the feScore/ghgScore columns of the same database.

Scores of −1 (EPA's code for "not rated") are treated as unavailable and displayed as "--".

Score labels are assigned as follows:

09 Derived Metrics

The following metrics are derived from the computed fuel cost data using standard arithmetic:

10 EPA Comparison Data

Two EPA comparison fields are available for context:

• EPA Annual Fuel Cost (fuelCost08): The EPA's own estimate of annual fuel cost for the vehicle, calculated assuming 15,000 miles per year at 55% city / 45% highway driving at current fuel prices. Displayed for comparison alongside the user's customized estimate.
• 5-Year Savings vs Average (youSaveSpend): EPA's estimate of how much the vehicle saves (positive) or costs more (negative) in fuel over 5 years compared to the average new vehicle. Directly from the EPA database.

Both fields are visualized together in Fig. 9 (see Visualizations), alongside the user's own parameter-based estimate, on three clearly distinct bases that are not meant to be read as three measurements of the same quantity.

11 Fuel Price Exposure

This section of the calculator shows how much the price series behind your vehicle's fuel has moved historically and how much of typical ownership cost is fuel. It replaces an earlier composite score. It contains six figures, a chart (Fig. 10) and a table, all computed from the price histories and the simulation (see Visualizations):

• Price series. The series that drives the vehicle's fuel cost: regular gasoline, diesel or residential electricity; plug-in hybrids use two.
• Standard deviation of annual price change. The population standard deviation of the year-over-year changes in that series over its full history.
• Simulated range against the flat estimate. The 10th and 90th percentiles of the simulated total cost in Fig. 9, each expressed as a percentage difference from the flat-price estimate.
• Fuel share of typical ownership cost. Fuel divided by fuel plus depreciation plus the other typical costs over your ownership period.
• Recent volatility. The population standard deviation of the latest ten annual changes in the series, shown next to the full-history value. Volatility changes over time, so a single full-history figure can understate or overstate present conditions.
• Fuel cost at risk. The 95th percentile of the simulated fuel cost over your ownership period (the same value shown as value at risk in Fig. 9e) minus the flat-price estimate, in dollars, and as a share of typical ownership cost. It is a plain percentile of the simulation: it carries no weights, scores or thresholds of the site's own.
• Fig. 10, your fuel bill by era. This figure is specific to your vehicle and inputs. For every first year of ownership that the data allows, the calculator prices the same driving (your vehicle, miles per year, driving mix and state, over your ownership period) at each year's real fuel price. The real price of a year is its nominal annual average divided by CPI-U (constant dollars) and taken relative to the latest year, so the bill for a window is your flat-price estimate rescaled by how the real price of that era compared with the latest year; a ratio of 1 reproduces the estimate. Plug-in hybrids scale their gasoline and electricity parts by their own series. The dashed line is your flat-price estimate, and the text under the figure gives the lowest and highest windows, the median across windows and the share of windows below your estimate. The figure and its text are shown only when at least three windows fit within the available price history (windows start no earlier than 1960, the first year of CPI-U in the source); otherwise the figure is omitted rather than shown empty. It is a replay of the past applied to today's driving, not a forecast.

  A small number of years are marked on the chart with a tick. Which years are marked is not a judgment call: they are, independently per fuel, among that fuel's own largest recorded annual price changes in the government series described under Price histories (for example regular gasoline's five largest moves are 2021, 1980, 1974, 1979 and 2022). The short label for each tick (for example "OPEC oil embargo"), listed just below the chart, is ordinary historical context for why that year moved, not a government-published figure and not itself sourced to a data file; only the year and the size of the price move are. Hovering a point on the chart names any marked year that falls inside that point's ownership window. Because each point on the curve is an average of the full ownership period starting there, a marked year's visible effect on the curve usually keeps building for a few years afterward rather than appearing as a spike exactly at the tick: a tick marks when a price moved, not when the curve peaks.

• Table. For each of the three series: years covered, mean, median and standard deviation of annual change, and the largest annual rise and fall with the year.

Known limits. Other liquid and gaseous fuels (E85, propane, natural gas) use the gasoline series as the nearest available history. The diesel series has 30 annual changes, the gasoline series 75 and the electricity series 65, so the diesel statistics rest on fewer observations.

12 Ownership Costs Beyond Fuel

Below the headline fuel estimate, the calculator shows a panel of typical other costs of owning a vehicle. These are national averages, not estimates for your vehicle, and are labelled that way on the page. Every number in the panel comes from a file in this project's data folder and is checked against that file by an automated script; the full list of figures, files and locations is at the end of this section.

What is included and where it comes from

• Insurance; maintenance and repairs; registration, licenses, leases and other charges; finance charges. U.S. Bureau of Labor Statistics (BLS), Consumer Expenditure Surveys, average annual expenditures of all consumer units (households), 2024. Each is divided by the BLS figure for average vehicles per consumer unit (1.9) to give a per-vehicle amount. Because both numbers are averages across all households, including those with no vehicle, the ratio is the average per vehicle across the whole fleet, including older vehicles and vehicles bought without a loan.
• Range across income groups. BLS Table 1101, 2022, gives the same items and the number of vehicles for each income fifth of households. The ratio of each fifth's per-vehicle spending to the all-household per-vehicle spending is applied to the 2024 average. The panel shows the lowest and highest of the five results. This shows how averages differed between income groups; it is not a range for your vehicle.
• Depreciation. BLS annual depreciation rates by automobile age (Monthly Labor Review, 2024, Chart 1 data, second quarter 1996 to first quarter 2022), compounded year by year over your ownership period, applied to the average price paid for a new vehicle (U.S. Bureau of Economic Analysis, NIPA Table 7.2.5S, average consumer expenditure per new car or per new light truck, mean of the latest twelve months in the file). Light trucks include pickups, sport utility vehicles and vans. You can replace the price with your own.

Assumptions and known limits

• The BLS insurance series has a break in 2019. That year BLS changed the source of its vehicle insurance estimate from its Diary Survey to its Interview Survey, having judged the Interview Survey a better measure (see BLS, Vehicle Insurance Spending in the 2019 Annual Calendar Year Estimates). Only 2019 and later insurance figures are used. As an independent check, the BLS Producer Price Index for private passenger auto insurance rose about 2.8% from 2018 to 2019, far less than the change in reported spending.
• The depreciation rates are in constant dollars. BLS estimated them from households' reported purchase prices, controlling for make, model and model year, after converting prices to constant 1982–84 dollars (see BLS, A consumption measure for automobiles). Applying them to current prices slightly overstates the loss in current dollars. Age 0 is a newly purchased automobile and its rate is used for the first year of ownership; BLS states that its data are not sufficient to estimate rates reliably for vehicles 11 years and older, and the age 11-and-over rate is used after year 11 only for very long ownership periods.
• These are averages across all vehicles and households. The panel does not adjust for vehicle type, fuel type, age, price, state, driver or coverage, except where you enter your own price or premium.
• Non-fuel amounts do not vary with your annual mileage in this estimate. Fuel does.
• Finance charges are an average over all vehicles, including those bought without a loan.
• Amounts are rounded to the nearest $10, except the fuel figure, which repeats the calculation above.

Not used

Data from private sources (for example AAA's annual cost study) is not used, because its terms of reuse could not be confirmed. Only U.S. government data is used.

13 How Your Mileage Compares

Below the ownership-costs panel, the calculator shows how the annual mileage you entered compares to the average resident of the state you selected (or the United States, if no state is selected): FHWA Table PS-1 state vehicle-miles-traveled divided by that same table's state population figure, the identical source and formula already used for this site's state pages.

What this figure is, and is not

• It is miles driven per resident of any age, including children and other non-drivers, not miles per licensed driver or per registered vehicle. Those are different, narrower figures published separately in How Many Miles Do Americans Drive (13,735 miles per licensed driver; 11,327 miles per registered vehicle, both 2024). The per-resident figure used here is necessarily lower than either, since its denominator includes people who don't drive at all.
• It is a state or national average, not a figure for people who share your specific age, income, commute, or household size. It does not adjust for any of those factors.

Not used

No private, state-level, or crowdsourced mileage or traffic data is used for this comparison, only the U.S. government's own FHWA Table PS-1 figures, the same source already cited above.

14 What Is Not Included

The headline estimate covers fuel and electricity only. The separate panel described in Ownership Costs Beyond Fuel adds national-average figures for other ownership costs; they are not specific to any vehicle. The following are not included anywhere on the page, because no per-vehicle U.S. government data source exists for them:

• Per-model depreciation, insurance, maintenance and financing: only national averages are shown (see the ownership-costs section); no federal agency publishes per-model figures.
• Tires: not shown separately (whether BLS counts tires within "maintenance and repairs" is not stated in the tables used).
• Sales tax: varies by state and locality; not included.
• Parking & tolls: vary by geography and individual usage; not included.

15 Limitations & Assumptions

Key assumptions and limitations:

• Constant fuel price in the primary estimate. The headline total applies today's fuel price to every year of ownership; there is no inflation modelling or price forecast in it. The historical replay (Fig. 3) and the simulation (Fig. 9) show how the estimate would change if past annual price changes recurred. They are backward-looking, they use nominal prices, and they are not forecasts.
• EPA test conditions. Label fuel economy is based on the EPA's 5-cycle laboratory procedure (city, highway, high-speed, air-conditioning and cold-temperature tests), which gives lower values than the basic 2-cycle city and highway tests; for the model year 2024 Toyota Prius, 57 MPG combined on the label versus 80 MPG from the 2-cycle tests (EPA Automotive Trends Report 2025, Table C.1). Real-world fuel economy still varies with driving style, speed, weather, terrain, cargo and vehicle condition.
• Constant annual mileage. Annual mileage is assumed constant across all years of ownership.
• No vehicle degradation. Fuel economy is assumed constant; no battery degradation (for EVs/PHEVs) or engine efficiency loss is modeled.
• Regional gas prices. FuelEconomy.gov's live price is a national average with no regional breakdown; selecting a state applies a real regional ratio on top of it (see Fuel Price Matching), but electricity prices are not regionally adjusted and remain a national figure.
• PHEV driving pattern. PHEV electric/gasoline split uses EPA's utility factors, which are population-level estimates, not your charging behavior. Your actual electric share depends on daily driving distance, charging frequency and route patterns.
• Ownership-costs panel. The non-fuel figures are national averages, not estimates for your vehicle; see the assumptions listed in Ownership Costs Beyond Fuel.
• Price series by fuel. Regular gasoline history stands in for other liquid or gaseous fuels; diesel and electricity use their own histories (see Fuel Price Exposure).
• Simulated price paths. The simulation resamples past annual price changes in runs of 3 years to keep the persistence found in the history; it does not model supply, demand, policy or inflation explicitly, and the simulated range is not a confidence interval. The assumption check in Visualizations shows how much the draw method matters. With diesel, 30 annual changes are available, so its range is the least certain.
• Nominal and real prices. The headline estimate, the replay and the simulation use nominal prices (dollars of the year in which the fuel is bought). Fig. 10, which prices your driving in each historical window, uses real prices, adjusted with CPI-U, and says so on the chart.
• Vehicles priced far from the average. Depreciation, insurance and the other ownership costs use national averages. For luxury, exotic or heavy-duty vehicles the average new-vehicle price understates the loss in value; enter your own price under Adjust for your car.
• Label fuel economy against real driving. No federal dataset in this project measures real-world fuel use for an individual model, so the estimate rests on EPA label values and has not been compared with measured driving for the same vehicle.

16 Data Freshness

Parts of the calculator update at different rates:

• Live on every page load: the vehicle picker, each vehicle's EPA record and the current national fuel and electricity prices from FuelEconomy.gov.
• Static, refreshed by hand: the three price histories (one new year each January or February, when EIA publishes the prior year's annual average), the model year 2026 reference population and reference prices (refreshed when a new model year is published), the state fuel prices (one week, refreshed periodically) and the BLS and BEA ownership-cost figures.

The live parts do not depend on the static parts; the static parts affect only how current the rankings, histories and ownership-cost averages are. A script (audit/run_all.py) rebuilds every embedded table from the files in the data folder and fails if the page differs.

17 Verification

Last reviewed October 31, 2026. How the numbers on the calculator are checked, and what the checks do not cover.

• Independent recomputation. An automated audit runs the real page in a browser and recomputes every displayed value from the raw EPA record and the fuel prices, without using the page's own code. The standard run uses a seeded sample of about 70 vehicles across gasoline, premium, diesel, hybrid, plug-in hybrid, electric, flex-fuel, natural gas and other fuel types, each in five setups that vary years, miles, driving mix, state and price overrides, plus vehicle comparisons. It makes more than 40,000 checks.
• The simulation is reproduced outside the page. The price simulation is seeded, so the audit repeats it bit for bit in a second program and compares every statistic the page prints. The tornado swings, the historical windows in Fig. 10, the ownership table, the rankings and the volatility statistics are recomputed the same way.
• The checks are themselves tested. Realistic bugs (a wrong operator, a swapped weight, a dropped year, a wrong constant) are planted one at a time in a copy of the page, and the audit must fail on each. Bugs that the audit missed led to new checks. A change that cannot alter any displayed value is listed in the script and not counted.
• Charts. Every chart is rendered at phone and desktop width and at extreme settings (0% and 100% city driving, 1 and 15 years, very low and very high mileage, each fuel type, comparison mode). Labels must be at least 11 px, have at least 4.5:1 contrast, not overlap, not be clipped and not be painted over.
• Not covered. Live values from FuelEconomy.gov are used as delivered and are not independently verified. Label fuel economy has not been compared with measured driving. Only one browser engine (Chrome) has been tested. The methodology has not been reviewed by anyone outside this project.