Outage Almanac

How we built this

What the numbers mean, what we removed, and where the records fall short.

The records

County figures come from the Event-correlated Outage Dataset in America, published by Pacific Northwest National Laboratory through the Open Energy Data Initiative. It is built from EAGLE-I outage data and lists county outages that reached at least 200 customers. We use 2015 to 2023. Population comes from the U.S. Census Bureau Vintage 2024 county estimates.

What we changed

  • We removed 10,547 of 1,371,873 records that looked like reporting faults: 10,317 brief spikes and 230 long records where the customer count barely moved.
  • We filtered and summarized the remaining records by county. The results are our own calculations, not figures published by the dataset authors.

What the terms mean

  • Customer-hours: customers out multiplied by hours out. 1,000 customers out for 2 hours is 2,000 customer-hours.
  • Outage time per resident: yearly customer-hours divided by county population. One account often serves several people, so this is a comparison measure and not hours per person. The US average is 2.44.
  • Major outage: an outage of at least 8 hours that reached customers equal to at least 1% of county residents.
  • Outage hours: how long at least 200 customers in the county were out. Most customers get power back sooner.
  • Peak customers: taken from records of at least 4 hours, because very short records are more likely to be faults.
  • Tiers: four equal groups of rated counties, split at 0.39, 0.81 and 1.61 times the US rate.

Data confidence and ratings

  • High: at least 7 years of records. Medium: at least 4 years.
  • Low (not rated): under 4 years of records, under 5,000 residents, or a peak above 80% of residents. We still show these pages, but we leave them out of rankings.
  • 2,656 of 2,966 counties are rated. Some larger counties are not rated because a reported peak looks too high for the population.

Limits

  • Records end in 2023, and many counties have fewer years of history because local utilities started reporting later.
  • Utilities that do not report, including some municipal systems and co-ops, can make a county look better than it is.
  • County totals do not show what happened on a particular street. The records do not include outage causes.
  • Connecticut, Puerto Rico and the US Virgin Islands are not included yet.
  • This is past data. It is not a forecast and not live outage information.

Credit and license

She, B., Adetola, V., & Yun, J. (2024). Event-correlated Outage Dataset in America. Open Energy Data Initiative, Pacific Northwest National Laboratory. Licensed under CC BY 4.0. Dataset funding: PNNL Laboratory Directed Research and Development program, Resilience Through Data-Driven, Intelligently Designed Control Initiative. This site is not endorsed by the authors or the U.S. Government.