Residential Electricity Use By State Data: Dataset Structure and Field Coverage

Four cataloged sources carry the query, and each owns a different slice of geography and grain. EIA Consumption & Efficiency - Electricity Surveys (RECS/CBECS/MECS) holds the household view: its RECS 2020 cycle sampled 18,496 households across 799 variables in a roughly 56 MB file, scaled to represent the 123.5 million U.S. housing units occupied as primary residences. It publishes state-level estimates for select states and rolls the remainder up to census division.

EIA State Energy Data System (SEDS) completes the map. It produces one annual estimate per state, per energy source and per sector for all 50 states, the District of Columbia, U.S. territories and a U.S. total, with consumption series reaching back to 1960.

Two supply-side companions report what utilities sold rather than how households lived. EIA Electric Power Annual devotes 14 Electricity Sales tables to sales, revenue, price and customer counts by sector and state, and EIA Electric Power Monthly refreshes state-level detail every month.

Sample design sets the limit. RECS 2020 drew 18,496 sampled households, and that sample supports published state estimates only for select states; everywhere else results roll up to census division, the DIVISION field in the microdata (observed value for household 100001: Mountain South).

Depth compensates for breadth. Published RECS cycles span 1978 to 2020 - 1993, 1997, 2001, 2005, 2009, 2015 and 2020, with a 2024 initial release listed - so the same state splits can be traced across four decades even where a single cycle stops at division level.

A repeatable build takes five moves:

1. Group on the finest geography that survives the full sample. Aggregate by the state postal code where identifiable and fall back to REGIONC and DIVISION elsewhere, so no household silently drops out of the national total you will be checked against.

1. Sum KWH and DOLLAREL, and keep the weather columns. Carrying HDD65 and CDD65 beside the totals lets a reader see whether a high-usage state ran hot or simply heated more.

1. Calibrate to the published estimate tables. EIA scales the sample to the 123.5-million-home universe, so raw sums drift from official figures wherever the design weights bite; reconcile your select-state rows to the published tables before publishing anything.

1. Extend and bridge. Add SEDS annual series for the states RECS does not publish, and fill the gap between survey waves with consumption_monthly.xlsx, the state-level monthly file running 2001 to present that sits behind EIA Electric Power Monthly.

Coverage first: all 50 states, the District of Columbia, U.S. territories and a U.S. total, with one annual estimate per state per energy source per sector, held to definitions meant to stay consistent over time and across sectors for analysis and forecasting. Consumption runs 1960-2024 and prices and expenditures from 1970.

The file layout rewards scripting. Every row keys on Data_Status (F final or P preliminary, prefixed by release year, so current values read 2024F), State (a two-character postal code, US for the national total) and MSN, a five-character series code followed by one column per year from 1960 to 2025. Codes_and_Descriptions.xlsx documents every code, and the consolidated Complete_SEDS.csv packs more than 2.5 million records into one long-format table.