Claim at v1.0.0

rlc-expansion-rearend

1e540f4845749f43f1fcb6cc1b637f04a1ee800fd04cccae708c980ed010505d

Data

{
  "has_outcome": true,
  "kind": "conflict",
  "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": "rlc-expansion-rearend",
  "prereg_branch": {
    "name": "prereg/rlc-expansion-rearend",
    "sha": "3e747066e417950621ada4cb0610650dbb5e77b4"
  },
  "prior": 0.3,
  "registered_at": "2026-11-13T12:39:17-0500",
  "registrant": "street-safety agent (Claude Opus 5.5), for J. Vernon",
  "registration": {
    "alternatives": [
      "Regression to the mean: new camera sites were chosen for high crash or violation counts; matching on pre-period crash counts and the two-year pre window reduce, but do not remove, this bias.",
      "Concurrent treatments at the same intersections in 2025-26 (street improvement projects, LPIs, daylighting, signal retiming); the SIP-excluded check addresses part of this.",
      "Traffic diverted from camera intersections to nearby streets; controls are at least 300 m from every camera, which limits but does not exclude spillover.",
      "A warning or test period before the first ticket, or tickets dated after the violation, shifting the true activation month.",
      "Changes in police crash reporting or geocoding in 2026 that differ between camera and control intersections.",
      "More traffic or congestion at camera sites in 2026 raising all crash types; the all-crash ratio is reported to show this."
    ],
    "graph_evidence": [
      {
        "findings": [
          {
            "abstract_check": "correct",
            "pubmed_id": 28454859,
            "statistic_type": "percent change",
            "statistic_value": -14,
            "variable_1_name": "activating red light cameras",
            "variable_2_name": "all fatal crashes at signalized intersections"
          },
          {
            "abstract_check": "correct",
            "pubmed_id": 28454859,
            "statistic_type": "percent change",
            "statistic_value": -21,
            "variable_1_name": "activating red light cameras",
            "variable_2_name": "fatal red light running crashes"
          },
          {
            "abstract_check": "wrong: abstract gives +16% for these 14 cities; +8% is for all 19",
            "pubmed_id": 28454859,
            "statistic_type": "percent change",
            "statistic_value": 8,
            "variable_1_name": "Removing red light cameras (2010-2014)",
            "variable_2_name": "all fatal crashes at signalized intersections"
          },
          {
            "abstract_check": "direction correct; value not in abstract",
            "pubmed_id": 32195995,
            "statistic_type": "coefficient",
            "statistic_value": -0.06,
            "variable_1_name": "Red light camera laws",
            "variable_2_name": "motor vehicle crash fatality rate"
          }
        ],
        "note": "Variable search (analyses/street_safety/graph_collect.py). Red-light camera associations come from two studies; none concerns rear-end or right-angle crashes. Abstracts checked: data/derived/street_safety/graph_check.json (13 of 21 relied-on records correct, 7 wrong, 1 unverifiable)."
      }
    ],
    "id": "rlc-expansion-rearend",
    "kind": "conflict",
    "literature": [
      {
        "citation": "Erke A (2009). Red light for red-light cameras? A meta-analysis of the effects of red-light cameras on crashes. Accid Anal Prev 41(5):897-905.",
        "doi": "10.1016/j.aap.2008.08.011",
        "finding": "In the best-controlled studies, rear-end crashes about +40% and right-angle crashes about -10%; all crashes about +15%; none significant. Results are more favourable without control for regression to the mean."
      },
      {
        "citation": "Hoye A (2013). Still red light for red light cameras? An update. Accid Anal Prev 55:77-89.",
        "doi": "10.1016/j.aap.2013.02.017",
        "finding": "Right-angle crashes -13%, rear-end +39%; right-angle injury crashes -33%, rear-end injury crashes +19%."
      },
      {
        "citation": "Hu W, Cicchino JB (2017). Effects of turning on and off red light cameras on fatal crashes in large U.S. cities. J Safety Res 61:141-148.",
        "doi": "10.1016/j.jsr.2017.02.019",
        "finding": "Fatal crashes at signalized intersections 14% lower after activation; 16% higher after removal (2010-2014 removals)."
      },
      {
        "citation": "Aeron-Thomas AS, Hess S (2005). Red-light cameras for the prevention of road traffic crashes. Cochrane Database Syst Rev CD003862.",
        "doi": "10.1002/14651858.CD003862.pub2",
        "finding": "Injury crashes RR 0.71 (0.55-0.92) in a small number of controlled before-after studies; rear-end effects inconsistent."
      },
      {
        "citation": "NYC DOT (2025-2026). NYC Red Light Camera Program addendum and press release of October 2026.",
        "doi": null,
        "finding": "At existing camera intersections, right-angle injury collisions -70% and rear-end injury collisions about -60% against a pre-installation average. No control group; this disagrees in sign with the meta-analyses on rear-end crashes."
      }
    ],
    "plan": "Analysis script analyses/street_safety/rlc_did.py at commit 793cc32 (B = 2000 bootstrap draws, seed 20261113; robustness rerun with B = 20000 and seeds 1-10). Treatment data: red-light camera intersections and the first month with tickets, from Parking Violations Issued FY2024-FY2027 (8zf9-spf8, m5vz-tzqv, 9mwx-gamw, pvqr-7yc4; violation code 7, sampled on the 1st and 15th of each month; analyses/street_safety/rlc_sites.py), geocoded to OpenStreetMap intersections and the LPI gazetteer (geocode_sites.py). Treated: intersections first ticketed January-August 2026 with no ticket in earlier sampled months. Post window: months a+1 to a+9 (a = first month), ending October 2026; pre window: the same calendar months in 2024 and 2025. Controls: 3 LPI intersections (xc4v-ntf4, LPI before 2024) per treated site, same borough, at least 300 m from any camera, matched on the pre-window count of all crashes. Crashes (h9gi-nx95, with vehicles bm4k-52h4) within 50 m. Estimator: ratio of ratios (treated post/pre over control post/pre); 95% percentile interval from a cluster bootstrap over matched sets. Secondary: placebo one year earlier; SIP sites (shr7-eqdc, 2024-2026) excluded. Expected discrimination before data (power.py, assumed rates): {\"n_treated\": 283, \"post_site_years\": 156.7}. Primary outcome: rear-end proxy count. Secondary (reported, not deciding): crashes citing 'Traffic Control Disregarded' (red-running proxy), predicted ratio 0.50-0.85.",
    "prediction": {
      "direction": "+",
      "magnitude": "Ratio of ratios for the rear-end proxy (crashes citing 'Following Too Closely', or with an in-traffic vehicle struck at 'Center Back End') between 1.10 and 1.60.",
      "refute_if": "The upper 95% bound is below 1.10 (the city data exclude the meta-analytic increase; a decrease as NYC DOT reports would fall here).",
      "rule8": {
        "expected_D_if_prediction_true": 5.86,
        "expected_post_events": 191,
        "expected_se_log": 0.117,
        "pred_log": 0.283
      },
      "support_if": "The 95% interval of the ratio of ratios lies above 1.00 and the test discriminates (rule 8: D >= 4 against the band's geometric centre, 1.33)."
    },
    "prior": 0.3,
    "registered_at": "2026-11-13T12:39:17-0500",
    "registrant": "street-safety agent (Claude Opus 5.5), for J. Vernon",
    "sha256": "a0d11a850d2da5029a3ae583a3ce37a4ff36a4b4846df639fa569276879ace9e",
    "statement": "At the intersections where New York City switched on red-light cameras in January-August 2026, rear-end crashes rose relative to matched signalized intersections, as the meta-analyses predict and contrary to NYC DOT's report of fewer rear-end injury crashes at its camera sites."
  },
  "registration_commit": {
    "commit": "3e747066e417950621ada4cb0610650dbb5e77b4",
    "committed_at": "2026-11-13T12:39:17-05:00"
  },
  "registration_file_sha256": "43b4b3b99083d4aad69e16798e120c427c8175bfb5f5ca7bafeb389c1ffca73f",
  "registration_path": "claims/registered/rlc-expansion-rearend.json",
  "registration_sha256": "a0d11a850d2da5029a3ae583a3ce37a4ff36a4b4846df639fa569276879ace9e",
  "registration_text": "{\n \"id\": \"rlc-expansion-rearend\",\n \"kind\": \"conflict\",\n \"statement\": \"At the intersections where New York City switched on red-light cameras in January-August 2026, rear-end crashes rose relative to matched signalized intersections, as the meta-analyses predict and contrary to NYC DOT's report of fewer rear-end injury crashes at its camera sites.\",\n \"graph_evidence\": [\n  {\n   \"note\": \"Variable search (analyses/street_safety/graph_collect.py). Red-light camera associations come from two studies; none concerns rear-end or right-angle crashes. Abstracts checked: data/derived/street_safety/graph_check.json (13 of 21 relied-on records correct, 7 wrong, 1 unverifiable).\",\n   \"findings\": [\n    {\n     \"pubmed_id\": 28454859,\n     \"variable_1_name\": \"activating red light cameras\",\n     \"variable_2_name\": \"all fatal crashes at signalized intersections\",\n     \"statistic_type\": \"percent change\",\n     \"statistic_value\": -14,\n     \"abstract_check\": \"correct\"\n    },\n    {\n     \"pubmed_id\": 28454859,\n     \"variable_1_name\": \"activating red light cameras\",\n     \"variable_2_name\": \"fatal red light running crashes\",\n     \"statistic_type\": \"percent change\",\n     \"statistic_value\": -21,\n     \"abstract_check\": \"correct\"\n    },\n    {\n     \"pubmed_id\": 28454859,\n     \"variable_1_name\": \"Removing red light cameras (2010-2014)\",\n     \"variable_2_name\": \"all fatal crashes at signalized intersections\",\n     \"statistic_type\": \"percent change\",\n     \"statistic_value\": 8,\n     \"abstract_check\": \"wrong: abstract gives +16% for these 14 cities; +8% is for all 19\"\n    },\n    {\n     \"pubmed_id\": 32195995,\n     \"variable_1_name\": \"Red light camera laws\",\n     \"variable_2_name\": \"motor vehicle crash fatality rate\",\n     \"statistic_type\": \"coefficient\",\n     \"statistic_value\": -0.06,\n     \"abstract_check\": \"direction correct; value not in abstract\"\n    }\n   ]\n  }\n ],\n \"literature\": [\n  {\n   \"citation\": \"Erke A (2009). Red light for red-light cameras? A meta-analysis of the effects of red-light cameras on crashes. Accid Anal Prev 41(5):897-905.\",\n   \"doi\": \"10.1016/j.aap.2008.08.011\",\n   \"finding\": \"In the best-controlled studies, rear-end crashes about +40% and right-angle crashes about -10%; all crashes about +15%; none significant. Results are more favourable without control for regression to the mean.\"\n  },\n  {\n   \"citation\": \"Hoye A (2013). Still red light for red light cameras? An update. Accid Anal Prev 55:77-89.\",\n   \"doi\": \"10.1016/j.aap.2013.02.017\",\n   \"finding\": \"Right-angle crashes -13%, rear-end +39%; right-angle injury crashes -33%, rear-end injury crashes +19%.\"\n  },\n  {\n   \"citation\": \"Hu W, Cicchino JB (2017). Effects of turning on and off red light cameras on fatal crashes in large U.S. cities. J Safety Res 61:141-148.\",\n   \"doi\": \"10.1016/j.jsr.2017.02.019\",\n   \"finding\": \"Fatal crashes at signalized intersections 14% lower after activation; 16% higher after removal (2010-2014 removals).\"\n  },\n  {\n   \"citation\": \"Aeron-Thomas AS, Hess S (2005). Red-light cameras for the prevention of road traffic crashes. Cochrane Database Syst Rev CD003862.\",\n   \"doi\": \"10.1002/14651858.CD003862.pub2\",\n   \"finding\": \"Injury crashes RR 0.71 (0.55-0.92) in a small number of controlled before-after studies; rear-end effects inconsistent.\"\n  },\n  {\n   \"citation\": \"NYC DOT (2025-2026). NYC Red Light Camera Program addendum and press release of October 2026.\",\n   \"doi\": null,\n   \"finding\": \"At existing camera intersections, right-angle injury collisions -70% and rear-end injury collisions about -60% against a pre-installation average. No control group; this disagrees in sign with the meta-analyses on rear-end crashes.\"\n  }\n ],\n \"prediction\": {\n  \"direction\": \"+\",\n  \"magnitude\": \"Ratio of ratios for the rear-end proxy (crashes citing 'Following Too Closely', or with an in-traffic vehicle struck at 'Center Back End') between 1.10 and 1.60.\",\n  \"support_if\": \"The 95% interval of the ratio of ratios lies above 1.00 and the test discriminates (rule 8: D >= 4 against the band's geometric centre, 1.33).\",\n  \"refute_if\": \"The upper 95% bound is below 1.10 (the city data exclude the meta-analytic increase; a decrease as NYC DOT reports would fall here).\",\n  \"rule8\": {\n   \"expected_post_events\": 191,\n   \"expected_se_log\": 0.117,\n   \"pred_log\": 0.283,\n   \"expected_D_if_prediction_true\": 5.86\n  }\n },\n \"plan\": \"Analysis script analyses/street_safety/rlc_did.py at commit 793cc32 (B = 2000 bootstrap draws, seed 20261113; robustness rerun with B = 20000 and seeds 1-10). Treatment data: red-light camera intersections and the first month with tickets, from Parking Violations Issued FY2024-FY2027 (8zf9-spf8, m5vz-tzqv, 9mwx-gamw, pvqr-7yc4; violation code 7, sampled on the 1st and 15th of each month; analyses/street_safety/rlc_sites.py), geocoded to OpenStreetMap intersections and the LPI gazetteer (geocode_sites.py). Treated: intersections first ticketed January-August 2026 with no ticket in earlier sampled months. Post window: months a+1 to a+9 (a = first month), ending October 2026; pre window: the same calendar months in 2024 and 2025. Controls: 3 LPI intersections (xc4v-ntf4, LPI before 2024) per treated site, same borough, at least 300 m from any camera, matched on the pre-window count of all crashes. Crashes (h9gi-nx95, with vehicles bm4k-52h4) within 50 m. Estimator: ratio of ratios (treated post/pre over control post/pre); 95% percentile interval from a cluster bootstrap over matched sets. Secondary: placebo one year earlier; SIP sites (shr7-eqdc, 2024-2026) excluded. Expected discrimination before data (power.py, assumed rates): {\\\"n_treated\\\": 283, \\\"post_site_years\\\": 156.7}. Primary outcome: rear-end proxy count. Secondary (reported, not deciding): crashes citing 'Traffic Control Disregarded' (red-running proxy), predicted ratio 0.50-0.85.\",\n \"alternatives\": [\n  \"Regression to the mean: new camera sites were chosen for high crash or violation counts; matching on pre-period crash counts and the two-year pre window reduce, but do not remove, this bias.\",\n  \"Concurrent treatments at the same intersections in 2025-26 (street improvement projects, LPIs, daylighting, signal retiming); the SIP-excluded check addresses part of this.\",\n  \"Traffic diverted from camera intersections to nearby streets; controls are at least 300 m from every camera, which limits but does not exclude spillover.\",\n  \"A warning or test period before the first ticket, or tickets dated after the violation, shifting the true activation month.\",\n  \"Changes in police crash reporting or geocoding in 2026 that differ between camera and control intersections.\",\n  \"More traffic or congestion at camera sites in 2026 raising all crash types; the all-crash ratio is reported to show this.\"\n ],\n \"prior\": 0.3,\n \"registrant\": \"street-safety agent (Claude Opus 5.5), for J. Vernon\",\n \"registered_at\": \"2026-11-13T12:39:17-0500\",\n \"sha256\": \"a0d11a850d2da5029a3ae583a3ce37a4ff36a4b4846df639fa569276879ace9e\"\n}",
  "replication": {
    "built_at": "2026-11-13T23:36:06-0500",
    "checked_at": "2026-11-13T23:36:10-0500",
    "decides": "support_if: 95% interval above 1.00 and rule 8 D >= 4 against 1.33",
    "deciding_number": "city_effect: ratio of ratios for rear-end proxy crashes within 50 m, 283 camera intersections against matched controls, cluster bootstrap 95% interval (B=2000, seed 20261113)",
    "inputs_copied": 3,
    "manifest": "replication/rlc-expansion-rearend/manifest.json",
    "manifest_sha256": "e94912e912430eddc943df48d8a5c4f6fbd0efd446ff6cf35b3943650b6a432a",
    "package": "replication/rlc-expansion-rearend/",
    "recorded": {
      "estimate": 1.3266,
      "hi": 1.6499,
      "lo": 1.061,
      "n": 283,
      "placebo_estimate": 1.2496,
      "placebo_hi": 1.488,
      "placebo_lo": 1.0519
    },
    "status": "MATCH",
    "status_from": "replication/rlc-expansion-rearend/verify_result.json",
    "tables_not_rebuilt": 0,
    "tolerance": {
      "estimate": 0.0001,
      "hi": 0.0001,
      "lo": 0.0001,
      "n": 0,
      "placebo_estimate": 0.0001,
      "placebo_hi": 0.0001,
      "placebo_lo": 0.0001
    }
  },
  "statement": "At the intersections where New York City switched on red-light cameras in January-August 2026, rear-end crashes rose relative to matched signalized intersections, as the meta-analyses predict and contrary to NYC DOT's report of fewer rear-end injury crashes at its camera sites."
}

History in this collection

  1. addedv1.0.0Nov 14, 2026this version