EIA Commercial Buildings Energy Consumption Survey (CBECS 2018)

CBECS is the U.S. Energy Information Administration's periodic census of the American commercial buildings stock, and the 2018 Building Characteristics release is its structural core. Rather than one flat export it ships as two connected artifacts: 22 tables of detailed crosstabulations covering building count, floorspace, workers, construction vintage, equipment and energy use - each cut by principal building activity, size band, age band, Census region and climate zone - and a public-use microdata file of 6,436 sampled buildings with roughly 1,250 variables apiece, delivered alongside its codebook and user guides. Together they weight up to the full universe: about 5.9 million commercial buildings.

For Office REITs work this is the denominator dataset. Broker reports quote vacancy rates and net absorption; CBECS supplies what the numerator is a share of - how many office buildings exist, how big they run, when they were built, how densely they are worked, and what they burn doing it.

Get a sample of this dataset and read real rows before you commit pipeline time.

The record carries three verified headline rows from Table B1, the summary-characteristics crosstab, exactly as published:

table                          : Table B1 - Summary Characteristics
row                           : All buildings
buildings_thousand            : 5918
total_floorspace_million_sqft : 96423
total_workers_thousand        : 85796
mean_sqft_per_building_thous  : 16.3

table                          : Table B1 - Summary Characteristics
row                           : Office
buildings_thousand            : 970
total_floorspace_million_sqft : 16662
total_workers_thousand        : 32843
mean_sqft_per_building_thous  : 17.2

table                          : Table B1 - Summary Characteristics
row                           : Office complex
buildings_thousand            : 141
total_floorspace_million_sqft : 3823
total_workers_thousand        : 8446
mean_sqft_per_building_thous  : 27.2

That middle row is the one Office REITs analysts pin to the wall: 970,000 office buildings - one in six of America's commercial stock by count - holding 16.7 billion square feet, which works out to about 17% of all commercial floorspace but roughly 38% of everyone working in it. Density is the business model. And the third row explains product tiering: purpose-built office complexes average 27,200 square feet against a 16,300-square-foot all-buildings mean, so class-A campuses and suburban medical-office strips live at opposite ends of the same table. A full sample carries the entire measure set, not just the summary lines.

Ten variable groups anchor the microdata spine, each defined in the accompanying codebook. They run from identifiers through physical plant to modeled end-use loads:

Beyond these anchors the full file opens up roughly 1,250 variables - HVAC configuration, lighting types, cooking and refrigeration equipment, employment bands, operating hours, plug-load detail and more. Fields whose definitions sit below the headline ten are folded under additional fields on request: flag them in the sample form and they arrive itemised, defined and exemplified.

Geography - United States national estimates, disaggregated only by the four Census regions (Northeast, Midwest, South, West) and their nine divisions, New England through Pacific. State-, metro- and market-level detail is not published; masking protects individual respondents. If your thesis turns on Austin versus Columbus, this is the wrong instrument - pair it with metro feeds such as the NCREIF Property Index office returns.

Temporal - the 2018 survey year, released in 2021 and revised in December 2022. Prior editions reach back to 1979, which makes CBECS one of the few series long enough to measure how American office stock aged into today's obsolescence debate. Treat any single edition as a fixed structural snapshot rather than a moving feed.

Granularity - one record per sampled building, 6,436 of them weighted to represent approximately 5.9 million buildings, plus the pretabulated crosstabs for anyone who wants the answer without touching record-level data. Anything below the building - tenant, suite, lease - sits outside the survey's design.

API, files, or your warehouse. Daily, weekly, or hourly.

• Sizing the office market honestly. 970,000 buildings and 16.7 billion square feet turn "how big is US office" from a broker adjective into a count - and put any single metro's inventory into national proportion.
• Benchmarking density and space standards. The published means - 17,200 square feet per office building, 32.8 million workers across them - calibrate square-feet-per-worker assumptions for hybrid-era right-sizing models rather than guessing them.
• Underwriting energy and retrofit exposure. Annual electricity and natural gas consumption in thousands of Btu, plus modeled office-equipment loads, give analysts a structural baseline for operating-expense and decarbonization-capex screens across vintages.
• Segmenting by vintage and format. Construction-year categories split the stock into age bands, separating pre-1980s towers from post-2000 product - the fault line most obsolescence theses are really about.
• Regional tilt analysis. Region and division cuts show where the floorspace actually sits, which matters when a REIT's Sunbelt weighting needs a national benchmark to be judged against.

Investors and quant researchers get the structural denominator: a fixed, citable picture of what the office stock physically is while prices move around it - see investors and quants workflows. Market researchers and consultants use the crosstabs as slide-ready national benchmarks that survive client scrutiny because every figure carries its survey year - more in market researchers workflows. Data scientists and ML engineers join the microdata's building-level features - vintage, size, hours, fuel mix - into valuation and risk models as slow-moving covariates that rarely need re-engineering; the pattern is sketched on data scientists workflows. Competitive-intel and product teams building proptech tools lean on the codebook-defined schema so their data dictionaries stop being guesses - see competitive intel and product teams workflows. Journalists and academics take the attribution-grade framing: nationally representative, methodology published, decades of prior editions for trend claims.

Notes worth reading alongside this page:

• Office REITs data hub - the pooled industry view this record sits inside, alongside ticker-level quotes, fund flows and property-sector peers.
• Best Office REITs datasets - where CBECS ranks in the pool and what beats it when freshness matters more than representativeness.
• The U.S. Energy Information Administration source profile - the publisher's wider catalog, including household and manufacturing energy counterparts.
• NCREIF Property Index - Office Returns - appraisal-based office performance to lay over CBECS's structural baseline.
• Yahoo Finance Quote Pages - Office REIT Tickers - daily price reality for BXP, SLG and VNO against a stock-of-buildings anchor.
• Financial Modeling Prep Developer API - REIT Financials - statement-level fundamentals when the question moves from buildings to balance sheets.
• CBECS glossary entry - the survey's vocabulary, defined once so rows read cleanly.
• Get a sample of this dataset - the fastest route to full rows, the extended variable list and the codebook-backed definitions.