Job Sector Diversity
How evenly the jobs located in a tract are spread across industry sectors, rather than concentrated in one.
On this page
The Census Bureau reports the jobs at each workplace broken out by NAICS sector — twenty of them, from "Agriculture, Forestry, Fishing and Hunting" through "Public Administration" (LODES Dataset Structure, Version 8.4). This family summarises that breakdown in one figure: whether a tract's jobs sit in many sectors or in one.
The summary is a standard concentration statistic, the Herfindahl-Hirschman index, applied to a tract's distribution of jobs across the twenty sectors and expressed so that a tract spread evenly across sectors sits at the high end and a tract where a single sector holds nearly every job sits at the low end. The index is a published statistical concept, used in the same form by competition authorities and by regional economists.
Concentration is what the measure is about. A tract dependent on one sector is exposed to whatever happens to it; a mixed base absorbs a sector-specific shock more easily. The measure carries no judgement about which sectors a tract holds — only how many. For the composition itself, see the per-sector industry families built on the same source.
Metrics in this family#
Job sector diversity
The diversity figure at the most recent year available for the tract.
Sector diversity trend, 5-year
The change in that figure between 2018 and 2023.
Sector diversity trend, 2-year
The change between 2021 and 2023. Both trends are reported as a level difference in the same units the figure itself uses, since the measure is a bounded index rather than a count.
Considerations#
The measure is sensitive to how few jobs a tract has. A tract with a handful of jobs will read as either very concentrated or very mixed almost by accident, and small changes in a small job base move the figure a long way. Diversity is most meaningful in tracts with substantial employment.
Tract boundaries are administrative, not economic. A large mixed employment district split across several tracts can leave each individual tract looking specialised while the district as a whole is not. The reverse also happens where one tract happens to span the edge of two different districts.
Twenty sectors is the resolution the Bureau publishes. Within one of them the work may be varied or narrow, and the measure cannot tell the difference.
Underlying counts are protected. The Bureau's methods "distort values by infusing noise at the employer and establishment levels and by weighting to control totals," with the same factor applied to a worksite over time (LODES Design and Methodology Report). Small sector cells are the least reliable part of the distribution.
Coverage is limited to unemployment insurance covered wage and salary jobs, so the self-employed, informal workers, the military and postal workers do not appear. Alaska has published no workplace data since 2016 and Michigan none since 2021, so trend windows ending later than those years carry no value there.
References#
- 1LODES Dataset Structure, Version 8.4https://lehd.ces.census.gov/data/lodes/LODES8/LODESTechDoc8.4.pdf
- 2LODES Design and Methodology Reporthttps://www2.census.gov/library/working-papers/2025/adrm/ces/CES-WP-25-52.pdf