How we tested this

The comparison

Before 2024, some utilities saw business customers' electricity use grow much faster than others. In each grid region we took the fastest-growing third (growth measured from 2019 to 2022, using only data published before 2024) and compared their household prices, from January 2024 to June 2026, with the other utilities in the same region. We allowed for gas prices and for changes that hit the whole region. "Fastest-growing" is relative to the region, not a fixed threshold; in the South the line between the two groups was small, around one percent growth in business use over three years.

Could it be chance?

We shuffled which utilities counted as "fast-growing" 9,999 times, only swapping within a region, and measured the gap each time. Then we asked how often chance alone would produce a gap as big as the one we saw. About 40 times in a hundred. That is common, so the gap is not distinguishable from chance.

How small a change we could see

Before running the test, we estimated the smallest change it would catch most of the time: about $5 a month (4.4%). A change of about $2 a month would have been caught only about 24 times in a hundred. So a missing gap is not evidence that nothing is happening: the plausible range from the actual data, about $7 less to $3 more, is the honest summary. If the two groups would have drifted apart somewhat anyway, that range widens to roughly $8 less to $5 more. The advance estimate used a simpler, faster version of the test, so it is a rough guide.

Correlation and causation

This is a comparison, not a controlled experiment. We can't assign business demand to some towns and not others, so the test leans on one assumption: without the post-2023 demand surge, the two groups' prices would have kept moving together, as they did before 2024 (our pre-2024 check found no sign of drift). Even if a gap had appeared, it would show that prices moved differently where demand grew fastest, not that any particular data centre or factory changed any particular bill. "Business demand" here includes data centres, but also factories, electrification and the rebound after COVID. Our check of whether this measure tracks data centres specifically came out mixed, so we don't attribute anything to data centres.

More limits, in short

• A utility serving several states counts once per state. Units in the same company may move together, which our uncertainty does not fully capture.
• Three utility mergers were combined into single units for the whole period. This was decided from usage data before any prices were compared, and is listed in the pre-registration's Appendix A2.
• EIA published no monthly data for part of 2017, and its 2025 and 2026 numbers are preliminary. We will re-run on final figures when they are published and report any change.
• The test covers only utilities that sell both power and delivery.

• Were the groups already drifting apart before 2024? No sign of it (chance of seeing a pattern like this if there were no drift: 0.90). A pass.
• Does a fake start date produce a fake gap? We pretended the change began in 2019, when nothing should have happened. No gap appeared (chance value 0.76). A pass.
• Does one state decide the result? We dropped Virginia (this removed only one small utility, so it is a weak check), then each of 31 states in turn. No single state produced a clear gap (the lowest chance value was 0.21). A pass; the state-by-state check is the one that does the real work.

Passing these checks means we found no sign the comparison is unreliable. They do not show that growth in business demand leaves bills untouched. Tests for pre-existing drift are weak when there are few utilities, so we also report a more cautious range in the downloadable results file.

These are reported for completeness, and none was distinguishable from zero. We do not claim prices fell.

Other reasonable ways to run it	Gap a month	Chance value
Start the 'after' period in mid-2025 instead of 2024	$4 less	0.13
Use the exact growth rate instead of 'top third'	not in dollars*	0.19
Add co-operative and municipal utilities	$1 less	0.63
Weight utilities by number of homes served	$1 less	0.75
Allow each utility its own seasonal pattern	$2 less	0.43
Use a different gas-price control	$2 less	0.45
Drop the gas-price control	$2 less	0.39
Use EIA's final numbers (not yet published)	not yet run

*The exact-growth-rate version measures something different (the price change per unit of business-use growth, not a gap between two groups), so it has no dollar figure and its size is not comparable to the others. "Chance value": how often reshuffling the groups would produce a gap this big. Small values would suggest a real gap; none here is small.