How the Solar Savings Index is calculated

The calculation

For each utility, the modelled annual offset is its residential rate applied to the lesser of two quantities: the average annual electricity a household in its territory consumes, and the annual output of a standard array at that location.

Taking the lesser of the two matters. A household that uses far more than a roof-sized array can generate cannot offset the excess, and treating its whole bill as addressable would overstate the case for solar. Conversely, where the array out-produces the household, the surplus is not counted as savings here, because what it is worth depends on the utility’s net metering terms rather than on its retail rate.

Sources

Rates and consumption

EIA-861 2024 final release, bulk XLSX (f8612024.zip), data year 2024. EIA publishes no residential price column, so the rate is derived as revenue divided by sales for the residential class. That makes it a full-year average across all of a utility’s residential customers — not a tariff, not a marginal rate, and not the rate on any one bill. Average household consumption is that utility’s residential sales divided by its residential customer count.

Modelled solar output

PVGIS v5.2 PVcalc, European Commission Joint Research Centre. For each state we model one reference array — 8 kWp, 20° tilt, due south, 14% system losses — at that state’s population-weighted centre of population as published by the U.S. Census Bureau. Using one array everywhere means the differences between states reflect location alone.

51 of the 52 jurisdictions are modelled from the National Solar Radiation Database, the same dataset the U.S. National Renewable Energy Laboratory’s own PV calculator reads. Alaska lies outside its coverage and is modelled from ERA5 reanalysis instead, which is coarser. Retrieved 2026-08-20.

Why savings are a range

Solar output varies from year to year with the weather. The production model reports that variation as a standard deviation, and the index shows the resulting band rather than a single number that would imply more precision than exists.

Some territories show one figure instead of a range. That is not a missing value: it means the modelled array already covers the household’s entire annual consumption, so even a poor production year still offsets the whole bill. Those are the places where solar economics are least sensitive to a bad year.

Where this is weakest

• It is not a quote. Every figure describes a modelled array in an average household in a service territory. Your roof size, pitch, orientation, shading, and household size all move the result, and none of them are in this data.
• One point per state. Solar output is modelled at a single representative location per state. States that span several climates contain real internal variation this cannot express, and a household far from its state's population centre will differ from the figure shown.
• Multi-state utilities are averaged. When a utility files in several states, its states are averaged without weighting, because the source data does not say how its customers are distributed among them.
• Averages hide spread. A utility’s residential class contains households that use very little and households that use a great deal. The average is not the median and is not you.
• Net metering is not modelled. What a utility pays for exported surplus varies widely and changes frequently. Nothing on the index depends on it, which is also why surplus production is excluded from the savings figure.
• No installation cost. The index reports what a system would offset, not what it would cost or how long it would take to pay back. Those depend on quotes, incentives, and financing.

Coverage and exclusions

1,195 utilities publish enough EIA-861 residential detail to be scored. Utilities serving fewer than 50,000 residential customers — 938 of them — are scored but not given their own page, because a page serves nobody when nobody is looking for that utility. Their figures are computed and held in the same dataset as the utilities that are listed; they are simply not given a page here.

Tier labels on the index reflect our publishing priority. They are an editorial decision about what to build first and say nothing about the quality of a utility.

Provenance

Scoring methodology utility-page-priority-v2. The underlying evidence is committed and content-addressed; the population file is 2026-08-19-utility-page-priority-v2.csv at SHA-256 122bc9e90a9aae85…, and the EIA source archive is pinned at 77ce49c60ac5a6ba….

A change to the weights, thresholds, or production model requires a new methodology version and a published comparison against the previous one.