Loading data...
A live, daily test of carbon-aware electricity pricing — and the CO₂ it would save — on every grid whose data is open.
What if electricity got cheaper when the grid runs clean and pricier when it runs dirty? Every day this site runs that idea on real electricity-grid data and measures how much CO₂ it would save compared with today's standard pricing, on every grid that publishes its hourly generation openly.
A study of retail electricity tariffs whose price follows the grid's carbon intensity. It is not carbon pricing in the emissions-trading sense (a price on CO₂ allowances), and it is not carbon-aware computing (scheduling batch jobs). It asks one question: if households paid less when the grid is clean, how much CO₂ would move?
Generation mix, load and wholesale prices come from grid operators (ENTSO-E, EIA, NESO/Elexon, CAISO, OpenElectricity). Carbon intensity is computed from the mix with life-cycle emission factors (IPCC AR5). For Switzerland it is consumption-based: imports from neighbours are blended in at their own intensity.
How people respond to the price. Consumption is reallocated across the day with a constant elasticity of substitution (σ = 0.14, the residential time-of-use estimate of Caves, Christensen & Herriges). It is bounded against what real tariff pilots have observed, but it is not calibrated to field data for these grids. This is the weakest link and the numbers should be read with that in mind.
Total daily consumption is held fixed. Load moves between hours, it is never removed, so the saving comes from timing alone, not from people using less. Each hour is also held within a comfort band of its normal level.
The hourly tariff follows carbon directly, so it can only move load toward cleaner hours. Carbon peak pricing sits on top of a time-of-use tariff, whose "peak" is a demand peak, not a carbon peak. On solar-heavy grids the evening peak is cleaner than the gas-fired night, so the time-of-use part pushes load into dirtier hours and the carbon spike does not always outweigh it. A negative number there is the tariff design showing, not an error.
The carbon intensity used is the average of the mix in each hour. The strictly correct figure for a consequential claim is the marginal intensity (the plant that responds to a change in demand), which is not available from open data in every region. Average is used everywhere, and stated here rather than hidden.
Percent says whether the mechanism works on a grid. Tonnes per day says how much carbon that is worth. They rank grids differently; a small clean grid can score a high percentage and still save less than a large one scoring a fraction of it.
The master's thesis behind this site reported a 19% reduction for Switzerland over 2021–2024. The live numbers are far lower, for three reasons: a cleaner and flatter Swiss grid; a corrected demand model (the thesis used an own-price elasticity where a substitution elasticity belongs, which overstates the response several-fold); and energy is now held constant, whereas the thesis counted some reduced consumption as saving. The thesis also used a third-party carbon feed (Electricity Maps); the live site computes carbon intensity itself from grid operators' generation data, except for Great Britain, where the system operator's own published intensity is used.
The demand model has been revised as it was reviewed, and each revision is stamped on its rows. Since v4, time-of-use periods and the four blocks follow each region's local clock; earlier versions applied them to UTC hours, which put the US and Australian evening peak in the wrong part of the day. The dashboard reports one version at a time and never mixes them. Earlier rows are kept in the dataset. A newer version is shown once it has accumulated enough days.
Everything is computed from open data. The full hourly record (carbon intensity, load, every tariff, every modelled demand) is published alongside the daily table, so every figure on this page can be recomputed. Some operators drop single hours; a gap of up to two hours is filled from its neighbours and flagged in that record, and a day with a longer gap is left out until the data is complete.
Loading…
A grid can be added when its operator publishes hourly generation by source openly, because carbon intensity has to be computed hour by hour. Annual or monthly national statistics are not enough. That is why China is not covered yet: it publishes national generation monthly, not hourly by source. The same holds for most of Africa outside South Africa. Coverage grows as open hourly data appears.