{
 "validation_errors": [],
 "doc_id": "jof_12530",
 "ontology_version": "1.0",
 "prompt_version": "1.0",
 "model": "claude-opus-5-5",
 "effort": "medium",
 "annotated_at": "2026-11-03T22:57:47+00:00",
 "source_sha256": "e7091a0d4b8826fccbb9cdfd5a1a80d2b93abaa741a5d2db98fb07459b1721af",
 "parser": "pymupdf-jof-heuristics-1.3",
 "payload_chars": 21303,
 "usage": {
  "input": 7476,
  "output": 2100,
  "cache_read": 3968,
  "cache_write": 0
 },
 "cost_usd": 0.072698,
 "request_id": "req_011CggpcpmpCGxdqD1gXthDB",
 "annotation": {
  "genre": "survey_or_address",
  "subfield": "financial_econometrics",
  "subfield_secondary": "research methodology / meta-science of publication bias in asset pricing",
  "primary_method": "other",
  "identification": {
   "estimand": "not applicable (methodological address); the target quantity is the posterior probability that the null is true given a reported p-value and prior odds",
   "strategy": "not applicable; conceptual argument, a hypothetical ticker-symbol data-mining example, and Bayesian conversion of p-values using minimum Bayes factors",
   "key_assumptions": [
    "prior odds on the null vs. alternative can be specified from economic plausibility",
    "minimum Bayes factor bounds the evidence against the null"
   ],
   "threats_addressed": [
    "multiple testing",
    "p-hacking",
    "publication bias / file drawer effect",
    "low prior probability of true effects (rare incidence)"
   ],
   "causal_language_strength": 2,
   "uses_causal_claim": false
  },
  "abstract_moves": {
   "question": {
    "present": true,
    "sentence_index": 4
   },
   "gap": {
    "present": true,
    "sentence_index": 3
   },
   "method": {
    "present": true,
    "sentence_index": 5
   },
   "results": {
    "present": true,
    "sentence_index": 1
   },
   "magnitude": {
    "present": false,
    "sentence_index": -1
   },
   "contribution": {
    "present": true,
    "sentence_index": 6
   }
  },
  "introduction": {
   "moves": [
    {
     "paragraph_index": 0,
     "move": "motivation"
    },
    {
     "paragraph_index": 0,
     "move": "tension_or_puzzle"
    },
    {
     "paragraph_index": 1,
     "move": "mechanism_or_theory_sketch"
    },
    {
     "paragraph_index": 2,
     "move": "related_literature_positioning"
    },
    {
     "paragraph_index": 2,
     "move": "policy_or_broader_implication"
    },
    {
     "paragraph_index": 2,
     "move": "roadmap"
    }
   ],
   "contribution_paragraph_index": 2,
   "contribution_types": [
    "new_method",
    "synthesis"
   ],
   "roadmap_present": true,
   "literature_positioning_location": "both"
  },
  "section_roles": [
   {
    "idx": 0,
    "role": "introduction"
   },
   {
    "idx": 1,
    "role": "literature"
   },
   {
    "idx": 2,
    "role": "other"
   },
   {
    "idx": 3,
    "role": "other"
   },
   {
    "idx": 4,
    "role": "literature"
   },
   {
    "idx": 5,
    "role": "theory"
   },
   {
    "idx": 6,
    "role": "other"
   },
   {
    "idx": 7,
    "role": "theory"
   },
   {
    "idx": 8,
    "role": "theory"
   },
   {
    "idx": 9,
    "role": "theory"
   },
   {
    "idx": 10,
    "role": "theory"
   },
   {
    "idx": 11,
    "role": "robustness"
   },
   {
    "idx": 12,
    "role": "other"
   },
   {
    "idx": 13,
    "role": "conclusion"
   }
  ],
  "results_presentation": {
   "economic_significance_discussed": false,
   "statistical_significance_language": "mixed",
   "magnitude_statements_in_intro": true
  },
  "robustness": {
   "location": "none_seen",
   "types": []
  },
  "tables": {
   "convention_notes": "Tables are illustrative computation grids. Rows are significance levels or z-scores/p-values, and columns are prior odds labeled from 'long shot' to 'even odds'. Notes define the MBF and SD-MBF formulas. One table re-expresses p-values from published papers as Bayesianized p-values.",
   "units_stated_in_notes": true,
   "standard_errors_description": "none"
  },
  "figures": {
   "uncertainty_shown": "no"
  },
  "conclusion": {
   "bounded_claims": true,
   "policy_implications": true,
   "open_questions": true,
   "restates_contribution": false
  },
  "hedging": {
   "level": "moderate",
   "typical_devices": [
    "modal verbs (may, likely)",
    "framing as a conversation starter",
    "disclaimer that past research quality is high",
    "extended quotations from authorities (Hubble, Russell)"
   ]
  },
  "claim_evidence": [
   {
    "claim": "Searching many specifications can produce a seemingly significant factor from economically meaningless sorting such as ticker letters.",
    "evidence_type": "descriptive_statistic",
    "evidence_ref": "Introduction ticker example (t=3.23)"
   },
   {
    "claim": "Published factor t-statistics cluster just above conventional thresholds, consistent with publication bias.",
    "evidence_type": "figure",
    "evidence_ref": "Figure 1, Figure 6 (HLZ 2016)"
   },
   {
    "claim": "Economics and business publish supportive results at rates close to psychology and above the physical sciences.",
    "evidence_type": "figure",
    "evidence_ref": "Figure 2 (Fanelli 2010)"
   },
   {
    "claim": "The share of positive results has risen over time across sciences.",
    "evidence_type": "figure",
    "evidence_ref": "Figure 3 (Fanelli 2012)"
   },
   {
    "claim": "With low prior odds, even strict significance thresholds leave a high probability that the null is true.",
    "evidence_type": "table",
    "evidence_ref": "Table I, Table II"
   },
   {
    "claim": "The required t-statistic threshold depends on the prior plausibility of the hypothesis.",
    "evidence_type": "table",
    "evidence_ref": "Table III"
   },
   {
    "claim": "Reported p-values in published factor papers translate into varied Bayesianized p-values depending on prior odds.",
    "evidence_type": "table",
    "evidence_ref": "Table IV"
   }
  ],
  "structural_summary": "A presidential address that opens with a provocative hypothetical data-mining example and then frames the problem as an agency conflict between editors and authors. It brings in meta-scientific evidence from other fields and reviews what p-values mean and their limitations. It then discusses p-hacking and the low base rates of true effects, and makes the case for Bayesian reasoning. As a compromise, it proposes minimum Bayes factors and illustrates them with tables and applications to published papers. It notes limitations, closes with research-culture recommendations, and ends with a reflective, quotation-rich conclusion.",
  "writing_profile_keywords": [
   "presidential address",
   "first-person voice",
   "meta-science",
   "p-hacking",
   "Bayesian inference",
   "minimum Bayes factor",
   "publication bias",
   "normative recommendations",
   "illustrative examples"
  ],
  "confidence": 0.8,
  "notes": "This is not an empirical paper with identification. The figures are mostly reproduced from other studies. Section headings idx 2–4 are artifacts of figure panel labels; idx 4 continues the discussion of Fanelli's explanations. Robustness checks are not applicable, and Section VIII discusses limitations of the method rather than reporting robustness tests."
 },
 "deterministic_checks": {
  "roadmap_regex": false,
  "contribution_regex_in_intro": true,
  "magnitude_in_abstract": true,
  "abstract_n_sentences": 16,
  "intro_n_paragraphs": 3,
  "intro_n_words": 600,
  "causal_words_in_abstract": 0,
  "hedge_words_in_abstract": 3,
  "stars_mentioned": false,
  "clustered_se_mentioned": false,
  "internet_appendix_mentioned": false,
  "section_roles_rule": {
   "0": "introduction",
   "1": "results",
   "2": "other",
   "3": "other",
   "4": "other",
   "5": "other",
   "6": "other",
   "7": "other",
   "8": "other",
   "9": "other",
   "10": "other",
   "11": "robustness",
   "12": "other",
   "13": "conclusion"
  }
 },
 "schema_valid": true,
 "code_version": "feb6c9e-dirty"
}