{
 "rules": "supported    the pre-registered criterion holds in the primary specification AND the\n               sign survives every cell of the robustness grid where one exists\n  tentative    the criterion holds in the primary specification only\n  unsupported  the criterion fails in the primary specification",
 "claims": [
  {
   "id": "H1",
   "claim": "The share of gender-signalled people who are signalled as women increases, 1900-1963.",
   "status": "tentative",
   "evidence": {
    "slope_pp_per_decade": 3.11,
    "ci": [
     1.41,
     4.81
    ],
    "early_share": 0.5532,
    "late_share": 0.683,
    "multiverse_share_positive": 0.999,
    "multiverse_cells": 900,
    "multiverse_slope_range": [
     -0.27,
     5.84
    ]
   },
   "script": "scripts/analyze_trends.py, scripts/robustness.py",
   "result": "results/trend_summary.parquet",
   "figure": "figures/fig03_female_share.png",
   "page": "/timeline/"
  },
  {
   "id": "H2",
   "claim": "Women's share of authority roles is below their share of all people in every period, and the gap narrows.",
   "status": "unsupported",
   "evidence": {
    "by_method": {
     "A": {
      "lower_in_every_year": true,
      "gap_trend_pp_per_decade": 0.01,
      "gap_trend_ci": [
       -0.9,
       0.93
      ],
      "gap_narrows": false
     },
     "B": {
      "lower_in_every_year": true,
      "gap_trend_pp_per_decade": -0.08,
      "gap_trend_ci": [
       -1.22,
       1.07
      ],
      "gap_narrows": false
     }
    },
    "authority_slope_by_method_same_entities": {
     "A": [
      3.37,
      1.02,
      5.72
     ],
     "B": [
      3.22,
      1.34,
      5.1
     ],
     "C@0.3": [
      1.41,
      0.18,
      2.64
     ],
     "C": [
      1.3,
      0.08,
      2.51
     ],
     "C@0.7": [
      1.23,
      0.27,
      2.18
     ],
     "D@0.3": [
      2.22,
      0.91,
      3.53
     ],
     "D": [
      2.0,
      0.8,
      3.19
     ],
     "D@0.7": [
      1.79,
      0.67,
      2.91
     ],
     "E": [
      1.35,
      -1.02,
      3.71
     ]
    },
    "multiverse_B_share_positive": 0.993
   },
   "script": "scripts/analyze_trends.py, scripts/analyze_methods.py",
   "result": "results/role_trends.parquet",
   "figure": "figures/fig04_roles_female_share.png",
   "page": "/roles/"
  },
  {
   "id": "H3",
   "claim": "Adjusting for topic and newspaper changes the 1900-1963 trend by at least 25% of the raw slope.",
   "status": "supported",
   "evidence": {
    "raw_slope": 3.11,
    "raw_ci": [
     -0.86,
     7.08
    ],
    "adjusted_slope": 0.47,
    "adjusted_ci": [
     -1.33,
     2.27
    ],
    "relative_change": 0.85,
    "decomposition": {
     "topic_label": {
      "total_pp": 12.98,
      "composition_pp": 5.5,
      "within_pp": 7.49
     },
     "section": {
      "total_pp": 12.98,
      "composition_pp": 5.5,
      "within_pp": 7.48
     },
     "paper_group": {
      "total_pp": 12.98,
      "composition_pp": -5.18,
      "within_pp": 18.16
     }
    }
   },
   "script": "scripts/analyze_composition.py",
   "result": "results/topic_adjusted.parquet",
   "figure": "figures/fig06_raw_vs_adjusted.png",
   "page": "/topics/"
  },
  {
   "id": "H4",
   "claim": "Role-measurement methods produce materially different trend effect sizes (ratio > 2 or sign disagreement for at least one headline role).",
   "status": "supported",
   "evidence": {
    "PUBLIC_OFFICE": {
     "slope_min": 0.33,
     "slope_max": 3.49,
     "ratio": 10.58,
     "signs_agree": true,
     "methods": "A,B,C,D"
    },
    "BUSINESS": {
     "slope_min": 1.56,
     "slope_max": 2.08,
     "ratio": 1.33,
     "signs_agree": true,
     "methods": "A,B,C,D"
    },
    "PROFESSIONAL": {
     "slope_min": 3.09,
     "slope_max": 4.18,
     "ratio": 1.35,
     "signs_agree": true,
     "methods": "A,B,C,D"
    },
    "CIVIC": {
     "slope_min": 2.17,
     "slope_max": 5.29,
     "ratio": 2.43,
     "signs_agree": true,
     "methods": "A,B,C,D"
    },
    "AUTHORITY": {
     "slope_min": 1.3,
     "slope_max": 3.37,
     "ratio": 2.6,
     "signs_agree": true,
     "methods": "A,B,C,D"
    }
   },
   "script": "scripts/analyze_methods.py",
   "result": "results/method_agreement.parquet",
   "figure": "figures/fig08_method_comparison.png",
   "page": "/methods/"
  },
  {
   "id": "H5",
   "claim": "The lexical method (A) yields a larger absolute trend than the context-aware methods (B, D).",
   "status": "tentative",
   "evidence": {
    "roles": {
     "AUTHORITY": {
      "A": 3.37,
      "B": 3.22,
      "D": 2.0,
      "lexical_largest": true
     },
     "BUSINESS": {
      "A": 1.56,
      "B": 2.08,
      "D": 1.76,
      "lexical_largest": false
     },
     "CIVIC": {
      "A": 4.23,
      "B": 5.29,
      "D": 3.61,
      "lexical_largest": false
     },
     "PROFESSIONAL": {
      "A": 3.59,
      "B": 3.24,
      "D": 3.09,
      "lexical_largest": true
     },
     "PUBLIC_OFFICE": {
      "A": 3.49,
      "B": 3.29,
      "D": 0.79,
      "lexical_largest": true
     }
    },
    "n_roles_lexical_largest": 3,
    "n_roles": 5
   },
   "script": "scripts/analyze_methods.py",
   "result": "results/method_agreement.parquet",
   "figure": "figures/fig08_method_comparison.png",
   "page": "/methods/"
  },
  {
   "id": "H6",
   "claim": "The UNKNOWN share changes over time, and the women's-share trend is sensitive to how UNKNOWN is handled.",
   "status": "tentative",
   "evidence": {
    "unknown_slope_pp_per_decade": -0.65,
    "unknown_ci": [
     -1.02,
     -0.27
    ],
    "unknown_share_first": 0.7733,
    "unknown_share_last": 0.7411,
    "slopes_by_handling": {
     "female_share_classified": [
      3.2,
      1.51,
      4.89
     ],
     "female_share_lower_bound": [
      1.19,
      0.68,
      1.71
     ],
     "female_share_upper_bound": [
      0.55,
      0.18,
      0.92
     ]
    },
    "signs_differ_across_handling": false
   },
   "script": "scripts/analyze_trends.py",
   "result": "results/extraction_summary.parquet",
   "figure": "figures/fig02_unknown_share.png",
   "page": "/timeline/"
  },
  {
   "id": "H7",
   "claim": "Among attributed quotations, women's share of speakers is lower than their share of gender-signalled people.",
   "status": "supported",
   "evidence": {
    "pooled_diff_pp": -44.73,
    "pooled_ci": [
     -45.4,
     -43.99
    ],
    "by_period": {
     "1900-15": [
      -34.82,
      -36.36,
      -33.78
     ],
     "1918-30": [
      -43.3,
      -44.94,
      -41.75
     ],
     "1933-45": [
      -46.36,
      -47.68,
      -44.69
     ],
     "1948-63": [
      -51.16,
      -52.41,
      -50.07
     ]
    }
   },
   "script": "scripts/analyze_trends.py",
   "result": "results/quote_trends.parquet",
   "figure": "figures/fig12_quotes.png",
   "page": "/quotes/"
  }
 ]
}