|boundary = census|
|a census-designated boundary delineating a statistical area, not necessarily observable on the ground|
|Used on these elements|
|Status: in use|
|Tools for this tag|
A census-designated boundary is not necessarily observable on the ground as a boundary, the way an administrative boundary might be, but a census agency has designated a statistical area within the boundary. The statistical area may correspond to a populated human settlement known by a particular name.
In the United States, census-designed places, which correspond to human settlements, were incorrectly imported as boundary=administrative admin_level=8 boundaries, just like municipal boundaries. While some mappers deleted these boundaries outright as unobservable abstractions, other mappers simply retagged them as boundary=census to keep them from being treated by data consumers as administrative boundaries. The latter approach allows geocoders to continue to find CDPs by name, which is helpful when the CDP corresponds to a commonly used place name. However, even if a CDP boundary is deleted, a commonly used place name can still be represented by a place=hamlet or place=village point. Many such points were imported from GNIS.
An area representation may be desirable in some cases. For example, Bethesda, Maryland, is bounded on some sides by incorporated areas, so the boundaries are implicitly observable on the ground, and a resident at the far reaches of the CDP may consider themselves to be unambiguously within the CDP. However, Bethesda is relatively exceptional among CDPs in this regard. Moreover, there are still differing definitions of Bethesda's extents among the Census Bureau, USGS, USPS, and other agencies, which illustrates the problem with treating a CDP boundary as a definitive place feature.
Data consumers should treat boundary=census with care. Most of these boundaries originated with the TIGER 2005 import. Later, in 2010, the U.S. Census Bureau overhauled CDPs with new criteria that better aligned CDPs to colloquial place names. Many pre-2010 CDPs are still present in OSM. Moreover, the shapes of a TIGER-imported CDP boundary are can be as inaccurate as any TIGER-imported boundary, with inexplicable jogs and wildly exaggerated protrusions. Renderers should avoid delineating CDPs as crisp boundaries, but some treatments like blurred lines or faint area shading may be justified. Search engines should avoid including CDP names in returned addresses, because CDP names do not necessarily correspond to USPS post office names.