Extracting the zoning variables

Answer the questions in reference/variables.json (the 21 measured variables of Bartik, Gupta and Milo, The Costs of Housing Regulation, with the paper's own wording) for exactly one government, from the zoning ordinance text you submit beside your answers. The questions come from the Pioneer Institute's Massachusetts housing-regulation survey as rephrased in the paper's Table A8/A9; answer them as written, for the whole jurisdiction, not for one district.

Rules

• Every answer cites the text. Each answered variable carries at least one citation { "section": "...", "quote": "..." } where quote is a verbatim passage of at least 15 characters from ordinance_text. The checker searches for the quote, whitespace-insensitive and case-insensitive, and refuses the Submission if any quote is missing. Choose the shortest passage that settles the answer.
• Polarity is the dataset's, not the question's. adus_banned is true when the answer to "are ADUs allowed in any district?" is no. multifamily_not_allowed, townhouses_not_allowed, no_mixed_use_buildings, no_conversion_to_multifamily, no_flexible_zoning_by_right and no_flexible_zoning_by_permit are likewise inverted. polarity in variables.json states each rule.
• Numbers carry the dataset's unit: lot sizes in square feet (convert acres: 1 acre = 43,560 sq ft), frontage in feet, waiting time in days, counts as integers. highest_res_min_lot_size and lowest_res_min_lot_size range over residential districts; mean_res_min_lot_size is their arithmetic mean across residential districts, so cite the table you averaged.
• Not determinable is an answer. When the ordinance does not settle a question, set value to null, not_determinable to true, and say in reason what you looked for. Do not guess from a neighbouring town or from general knowledge.
• The text you cite is the text you submit. ordinance_text is the plain-text or HTML rendering of the ordinance you read (the publisher's HTML, or the text layer of the PDF). Record the original's URL, retrieval date and SHA-256 under source so a reviewer can fetch the same bytes.

Shape of extraction

{
  "government": { "census_id_pid6": "100019", "name": "CITY OF PRATTVILLE", "state": "AL" },
  "source": { "url": "https://...", "retrieved_at": "2026-10-16", "original_sha256": "..." },
  "answers": {
    "adus_banned": { "value": false, "citations": [ { "section": "§ 78-402(b)", "quote": "Accessory dwelling units are permitted by special exception in the R-1 and R-2 districts" } ] },
    "highest_res_min_lot_size": { "value": 15000, "unit": "sq_ft", "citations": [ { "section": "Table 78-5", "quote": "R-1 Minimum lot area 15,000 square feet" } ] },
    "max_review_waiting_time": { "value": null, "not_determinable": true, "reason": "No review timelines are stated for site plan or special exception.", "citations": [] }
  }
}

All 21 keys must be present under answers.

Auditing a published row

When the bound government input carries published_row (the values the American Zoning Project's model produced), the task is the same extraction; the checker additionally reports, per variable, whether your answer agrees with the published value, and the reviewer weighs disagreements. Do not copy the published value; answer from the text and let the comparison fall where it falls. A documented disagreement with a citation is the most valuable outcome of an audit.