LEHD Origin-Destination Employment Statistics (LODES), Workplace Area Characteristics

LODES is released in three forms. The file used here, Workplace Area Characteristics, totals "jobs by work Census Block"; the other two total jobs by home block, or by both (LODES Dataset Structure, Version 8.4).

Everything on this page therefore counts workplaces. A tract with a large office park has a high job count even if few of its residents work there, and a residential tract a low one even if every adult in it is employed. Read these as a map of where employment sits, not as a measure of whether local people have work.

LODES counts "unemployment insurance covered wage and salary jobs, as reported by state labor market information offices" and, from 2010, by the Office of Personnel Management. State data covers private, state and local government employers and accounts for "approximately 97 percent of civilian wage and salary jobs, as estimated by the Bureau of Labor Statistics." Outside that universe are "the military and other security-related federal agencies, postal workers, some employees at nonprofits and religious institutions, informal workers, and the self-employed" (LODES Design and Methodology Report). A job counts if a person had positive earnings from a firm in both the April–June reference quarter and the quarter before it.

Jobs are also published by industry. The file carries a count of jobs for each of twenty NAICS sectors — from "NAICS sector 11 (Agriculture, Forestry, Fishing and Hunting)" through "NAICS sector 92 (Public Administration)" (LODES Dataset Structure, Version 8.4). That breakout is what the twenty industry families below are built on; the sector definitions themselves belong to the North American Industry Classification System.

The Bureau publishes 2002 onward. Birdseye uses 2015 onward, through the most recent release. That is not the same as today: LODES appears roughly a year and a half after the period it describes.

Two gaps matter. Alaska has no workplace data from 2017 on, and Michigan none for 2022 and 2023. Where a state stopped publishing, we show its most recent available year rather than a false decline.

These counts are deliberately imprecise. The Bureau states that its disclosure avoidance methods "distort values by infusing noise at the employer and establishment levels and by weighting to control totals." The noise is "dynamically consistent," the same factor applied to a worksite over time "so that longitudinal changes in employment are not distorted." Small counts are the least reliable; changes over time are better protected than levels.

The Bureau publishes at the 2020 census block, on 2024 TIGER/Line boundaries. Birdseye reports at census tract, which contains blocks exactly, so no estimation is involved. Around 1,170 of the roughly 85,184 tracts we map carry no workplace record.

Workplace address is what the employer reported, which "may or may not be the actual location that a worker reports to most often" — a real issue for construction and for multi-site employers filing a single address. LODES says nothing about telework or commute mode.

Birdseye figures derived from this source are our own estimates. The Census Bureau is cited as the source of the original data only and is not responsible for our calculations.