Claim at v1.0.0

construction-aerial-triangulation

cca98f8eaaa8f8723ef213222156f8abb847bd43cd6a6093e5e318b74ec537cf

Data

{
  "has_outcome": true,
  "kind": "bridge",
  "lineage": {
    "audit": "docs/lineage_audit.md; data/derived/lineage_audit.json",
    "class": "none",
    "class_meaning": "the audit found no city-data source fetched before registration",
    "classes": [],
    "early_sources": []
  },
  "outcome": "construction-aerial-triangulation",
  "prior": 0.45,
  "registered_at": "2026-11-12T15:14:19-0500",
  "registrant": "Claude (construction agent), for Jeff Vernon",
  "registration": {
    "alternatives": [
      "Seasonal and sun-angle differences in NAIP make unchanged lots look changed",
      "Tall neighbouring buildings lean over small lots in an orthophoto and hide them",
      "Interior gut renovations under an NB filing (rare) or enlargements produce little roof change",
      "Demolition alone (lot cleared, not rebuilt) counts as visible change but is not a completed building"
    ],
    "chain": [
      "construction-permit-funnel",
      "construction-four-year-completion"
    ],
    "graph_evidence": [
      {
        "concepts": [
          "population growth",
          "construction"
        ],
        "direction": "+",
        "n_findings": 1,
        "n_studies": 1,
        "pubmed_ids": [],
        "relationship_id": "B0QiEej1E1|COEFxSZVXO",
        "summary": "Population growth rate and expansion rate of construction land, Chongqing over 20 years (coefficient 1.82; doi:10.1117/12.815931). The only construction-supply finding found in the graph."
      },
      {
        "association_id": "aFGrNWb1i5",
        "concepts": [
          "housing (vacant housing demolitions)",
          "crime"
        ],
        "direction": "-",
        "n_findings": 1,
        "n_studies": 1,
        "pubmed_ids": [
          36228876
        ],
        "summary": "Vacant housing demolitions and change in crime, mean difference -0.068 (CI -0.119 to -0.018), PMID 36228876. Shows the graph treats land and buildings as exposures for health and crime, not as supply outcomes."
      },
      {
        "association_id": "(search)",
        "concepts": [
          "construction",
          "permit",
          "completion",
          "vacant",
          "demolition",
          "renovation"
        ],
        "direction": "0",
        "n_findings": 0,
        "n_studies": 0,
        "pubmed_ids": [],
        "summary": "Variable and concept searches (13 terms, data/derived/construction/graph_search.json) found no finding on building permits, permit-to-completion ratios, lags or permit lapse. The construction concept (COEFxSZVXO) has 3 relationships in all (wage, population growth, natural disasters)."
      }
    ],
    "id": "construction-aerial-triangulation",
    "kind": "bridge",
    "literature": [
      {
        "citation": "NYC Department of City Planning, 'New Housing Permits in the First Half of 2022' (information brief, 2022)",
        "doi": null,
        "finding": "'Typically, 80 to 90 percent of permitted projects are completed within four years.' DOB issued permits for about 58,600 homes in the first half of 2022, ahead of the 421-a expiry.",
        "url": "https://atlanticyardsreport.blogspot.com/2022/11/dcp-expiration-of-421-tax-break-drove.html (quotes the brief)"
      },
      {
        "citation": "NYU Furman Center, State of New York City's Housing and Neighborhoods / 2019 data on the housing stock",
        "doi": null,
        "finding": "Median time from new-building permit to certificate of occupancy was about 1.9 years for buildings permitted in 2009-2014.",
        "url": "https://furmancenter.org/thestoop/entry/2019-data-on-new-york-citys-housing-stock"
      }
    ],
    "plan": "Sample NB jobs from ipu4-2q9a (first NB permit 2018-2020) and DOB NOW (w9ak-ipjd New Building with first_permit_date 2018-2020), stratified by nothing, seed fixed. Lot polygon from PLUTO (64uk-42ks geom) by BBL. NAIP tiles from the Planetary Computer STAC for the lot box plus a 15 m margin, earlier and later years. Primary measure: my visual judgement of each before/after pair (changed / unchanged / unclear), made on a label-free grid BEFORE the CO status is joined, saved as data/derived/construction/aerial_judgements.csv. Secondary measure: mean absolute difference of grey levels inside the lot polygon after per-image standardisation, with a threshold set by eye on 6 cases. Then join CO status and compute the 2x2 table and kappa. Figure: grid of before/after thumbnails.",
    "prediction": {
      "direction": "+",
      "magnitude": "Random sample of 48 NB jobs (seed 20261112) first permitted 2018-2020. Visible change of the lot between the latest NAIP year <= 2017 and the latest year >= 2022: 65% to 90% of jobs. Agreement of visible change with 'has a CO by the later image date': Cohen's kappa 0.2 to 0.7. Jobs changed on the aerial but without a CO: 10% to 35%. Jobs with a CO but no visible change: under 10%.",
      "refute_if": "visible-change share < 0.45, or CO-but-no-change share >= 0.25 (the aerial and the CO disagree about what was built)",
      "support_if": "visible-change share >= 0.6 AND kappa > 0.2 AND CO-but-no-change share < 0.15"
    },
    "prior": 0.45,
    "registered_at": "2026-11-12T15:14:19-0500",
    "registrant": "Claude (construction agent), for Jeff Vernon",
    "sha256": "b42be119e44d67ed12a7a931fdb8041c865bbcf97b3332de4bb3df303a9738e6",
    "statement": "Aerial photographs (USDA NAIP) give a third, independent record of whether permitted new buildings were built: for a random sample of NB jobs first permitted 2018-2020, visible change on the lot between a pre-permit and a recent NAIP year agrees with the CO record, and shows sites under construction that the CO record misses."
  },
  "registration_commit": {
    "commit": "2a0a43902986996647ae72174b3ecc906e59d2e1",
    "committed_at": "2026-11-12T15:14:20-05:00"
  },
  "registration_file_sha256": "bc44504d7ae4352b06f40b867ff60f49472aa8409266be8bb7fb5959446313a3",
  "registration_path": "claims/registered/construction-aerial-triangulation.json",
  "registration_sha256": "b42be119e44d67ed12a7a931fdb8041c865bbcf97b3332de4bb3df303a9738e6",
  "registration_text": "{\n \"id\": \"construction-aerial-triangulation\",\n \"kind\": \"bridge\",\n \"statement\": \"Aerial photographs (USDA NAIP) give a third, independent record of whether permitted new buildings were built: for a random sample of NB jobs first permitted 2018-2020, visible change on the lot between a pre-permit and a recent NAIP year agrees with the CO record, and shows sites under construction that the CO record misses.\",\n \"graph_evidence\": [\n  {\n   \"relationship_id\": \"B0QiEej1E1|COEFxSZVXO\",\n   \"concepts\": [\n    \"population growth\",\n    \"construction\"\n   ],\n   \"n_findings\": 1,\n   \"n_studies\": 1,\n   \"direction\": \"+\",\n   \"summary\": \"Population growth rate and expansion rate of construction land, Chongqing over 20 years (coefficient 1.82; doi:10.1117/12.815931). The only construction-supply finding found in the graph.\",\n   \"pubmed_ids\": []\n  },\n  {\n   \"association_id\": \"aFGrNWb1i5\",\n   \"concepts\": [\n    \"housing (vacant housing demolitions)\",\n    \"crime\"\n   ],\n   \"n_findings\": 1,\n   \"n_studies\": 1,\n   \"direction\": \"-\",\n   \"summary\": \"Vacant housing demolitions and change in crime, mean difference -0.068 (CI -0.119 to -0.018), PMID 36228876. Shows the graph treats land and buildings as exposures for health and crime, not as supply outcomes.\",\n   \"pubmed_ids\": [\n    36228876\n   ]\n  },\n  {\n   \"association_id\": \"(search)\",\n   \"concepts\": [\n    \"construction\",\n    \"permit\",\n    \"completion\",\n    \"vacant\",\n    \"demolition\",\n    \"renovation\"\n   ],\n   \"n_findings\": 0,\n   \"n_studies\": 0,\n   \"direction\": \"0\",\n   \"summary\": \"Variable and concept searches (13 terms, data/derived/construction/graph_search.json) found no finding on building permits, permit-to-completion ratios, lags or permit lapse. The construction concept (COEFxSZVXO) has 3 relationships in all (wage, population growth, natural disasters).\",\n   \"pubmed_ids\": []\n  }\n ],\n \"literature\": [\n  {\n   \"citation\": \"NYC Department of City Planning, 'New Housing Permits in the First Half of 2022' (information brief, 2022)\",\n   \"doi\": null,\n   \"url\": \"https://atlanticyardsreport.blogspot.com/2022/11/dcp-expiration-of-421-tax-break-drove.html (quotes the brief)\",\n   \"finding\": \"'Typically, 80 to 90 percent of permitted projects are completed within four years.' DOB issued permits for about 58,600 homes in the first half of 2022, ahead of the 421-a expiry.\"\n  },\n  {\n   \"citation\": \"NYU Furman Center, State of New York City's Housing and Neighborhoods / 2019 data on the housing stock\",\n   \"doi\": null,\n   \"url\": \"https://furmancenter.org/thestoop/entry/2019-data-on-new-york-citys-housing-stock\",\n   \"finding\": \"Median time from new-building permit to certificate of occupancy was about 1.9 years for buildings permitted in 2009-2014.\"\n  }\n ],\n \"prediction\": {\n  \"direction\": \"+\",\n  \"magnitude\": \"Random sample of 48 NB jobs (seed 20261112) first permitted 2018-2020. Visible change of the lot between the latest NAIP year <= 2017 and the latest year >= 2022: 65% to 90% of jobs. Agreement of visible change with 'has a CO by the later image date': Cohen's kappa 0.2 to 0.7. Jobs changed on the aerial but without a CO: 10% to 35%. Jobs with a CO but no visible change: under 10%.\",\n  \"support_if\": \"visible-change share >= 0.6 AND kappa > 0.2 AND CO-but-no-change share < 0.15\",\n  \"refute_if\": \"visible-change share < 0.45, or CO-but-no-change share >= 0.25 (the aerial and the CO disagree about what was built)\"\n },\n \"plan\": \"Sample NB jobs from ipu4-2q9a (first NB permit 2018-2020) and DOB NOW (w9ak-ipjd New Building with first_permit_date 2018-2020), stratified by nothing, seed fixed. Lot polygon from PLUTO (64uk-42ks geom) by BBL. NAIP tiles from the Planetary Computer STAC for the lot box plus a 15 m margin, earlier and later years. Primary measure: my visual judgement of each before/after pair (changed / unchanged / unclear), made on a label-free grid BEFORE the CO status is joined, saved as data/derived/construction/aerial_judgements.csv. Secondary measure: mean absolute difference of grey levels inside the lot polygon after per-image standardisation, with a threshold set by eye on 6 cases. Then join CO status and compute the 2x2 table and kappa. Figure: grid of before/after thumbnails.\",\n \"alternatives\": [\n  \"Seasonal and sun-angle differences in NAIP make unchanged lots look changed\",\n  \"Tall neighbouring buildings lean over small lots in an orthophoto and hide them\",\n  \"Interior gut renovations under an NB filing (rare) or enlargements produce little roof change\",\n  \"Demolition alone (lot cleared, not rebuilt) counts as visible change but is not a completed building\"\n ],\n \"prior\": 0.45,\n \"registrant\": \"Claude (construction agent), for Jeff Vernon\",\n \"chain\": [\n  \"construction-permit-funnel\",\n  \"construction-four-year-completion\"\n ],\n \"registered_at\": \"2026-11-12T15:14:19-0500\",\n \"sha256\": \"b42be119e44d67ed12a7a931fdb8041c865bbcf97b3332de4bb3df303a9738e6\"\n}",
  "statement": "Aerial photographs (USDA NAIP) give a third, independent record of whether permitted new buildings were built: for a random sample of NB jobs first permitted 2018-2020, visible change on the lot between a pre-permit and a recent NAIP year agrees with the CO record, and shows sites under construction that the CO record misses."
}

History in this collection

  1. addedv1.0.0Nov 14, 2026this version