Five DINS fields measured as numbers publish a zero nobody has ruled on, four of them above 94 percent

[ChelseaKR]

What the survey found

python -m perimeter.survey (added in #82) run over the pinned 2026-08-07 retrieval reports five DINS fields measured as numbers whose recorded zeros nobody has ruled on:

Field	Zeros	Of numbers	Share	Empty	Declared
NOOFCARSONPROPERTY	55,831	56,188	99.4%	76,334	nothing (Basis.NONE, no note)
NOOUTBUILDINGSDAMAGED	59,628	61,939	96.3%	70,583	nothing
NOOUTBUILDINGSNOTDAMAGED	58,888	61,958	95.0%	70,564	nothing
NUMBEROFUNITPERSTRUCTURE	58,411	61,929	94.3%	70,593	nothing
ASSESSEDIMPROVEDVALUE	6,613	126,482	5.2%	6,040	nothing

Each is FieldSpec(..., numeric=True) with no unknown_markers, no recorded_absences, and Basis.NONE. So a 0 in any of them is currently classified present, counted as a recorded value, and published in dins-coverage.json and on dins.html as a field the agency filled in.

Why this is the project's own question

This is ADR-0006 (a zero is a judgment call) in the shape the project has already met once. YEARBUILT held 12,148 literal zeros, and #27 decided that no structure standing in a California wildfire was built in year 0, so a zero there is the parcel record carrying no year rather than a recorded construction year. Counting it as one published an absence as a value.

NUMBEROFUNITPERSTRUCTURE is the same sentence with a different noun. 58,411 structures recorded as containing zero units is not a plausible reading of a count of units; a structure with no units is not a structure. NOOFCARSONPROPERTY at 99.4% is starker still: on that reading, 357 of the 56,188 inspections that recorded anything at all found a car.

The two outbuilding counts are the interesting pair, because for them zero is genuinely plausible. Most inspected parcels really may have no damaged outbuildings, and 0 is then a finding, not an absence. That is exactly why this needs a person: the same numeric shape carries opposite meanings across four adjacent columns, and the survey can measure the distribution but cannot read the form.

ASSESSEDIMPROVEDVALUE is a third case again: a parcel genuinely assessed at zero improved value is possible (bare land, an exempt parcel), and 5.2% is a plausible rate for it.

What is needed, and from whom

A reviewer's reading of the DINS inspection form and Database Dictionary, per field, deciding for each of the five whether a recorded 0 is:

1. a measurement of zero (leave it present, and say in docs/MARKERS.md that it was reviewed and left), or
2. a marker for a field the inspector did not complete (declare unknown_markers={"0"} with Basis.INFERRED and a note, as YEARBUILT has), or
3. undecidable from the published documentation (which is itself a finding worth a note, and a candidate for the CEC-style domain request this portfolio already uses elsewhere).

Nothing has been declared, and nothing in #82 moves an artifact or a page. The survey reports; the reading is yours.

Reproducing

make survey    # over data/raw/, after make acquire

The relevant rows are recorded_zeros.undeclared_zeros in survey.json, and the same fields appear under their own headings in survey.md with basis: unreviewed.

One check on the tool's arithmetic: the same run reproduces YEARBUILT's 12,148 zeros, the figure section 7 of docs/MARKERS.md already records, and correctly reports it as declared rather than open.

Prepared with AI assistance; reviewed before submission.