{
 "validation_errors": [],
 "doc_id": "jof_12898",
 "ontology_version": "1.0",
 "prompt_version": "1.0",
 "model": "claude-opus-5-5",
 "effort": "medium",
 "annotated_at": "2026-11-03T23:14:31+00:00",
 "source_sha256": "f45750a9f65a52b60002b46774a4e98c9a04c68d6aa41d1b7f4e131ab34a3713",
 "parser": "pymupdf-jof-heuristics-1.3",
 "payload_chars": 19638,
 "usage": {
  "input": 6825,
  "output": 1911,
  "cache_read": 3968,
  "cache_write": 0
 },
 "cost_usd": 0.066314,
 "request_id": "req_011CggqtyUFnrkgoaGZqPnyW",
 "annotation": {
  "genre": "empirical",
  "subfield": "asset_pricing",
  "subfield_secondary": "financial_econometrics",
  "primary_method": "other",
  "identification": {
   "estimand": "Latent continuous and jump factor structure (number of factors, loadings, explained variation) of individual stock returns, and intraday vs overnight risk premia of these factors",
   "strategy": "PCA on locally estimated continuous and jump covariance matrices from five-minute high-frequency returns, with truncation to separate jumps; comparison with observable proxy factors via generalized correlations",
   "key_assumptions": [
    "approximate local factor model with locally constant loadings",
    "truncation threshold correctly separates continuous and jump increments",
    "five-minute sampling limits microstructure noise and asynchronicity"
   ],
   "threats_addressed": [
    "microstructure noise and asynchronous trading",
    "factor misspecification from prespecified factors",
    "dependence on choice of characteristics",
    "panel composition (balanced vs unbalanced)"
   ],
   "causal_language_strength": 1,
   "uses_causal_claim": false
  },
  "abstract_moves": {
   "question": {
    "present": true,
    "sentence_index": 0
   },
   "gap": {
    "present": false,
    "sentence_index": -1
   },
   "method": {
    "present": true,
    "sentence_index": 1
   },
   "results": {
    "present": true,
    "sentence_index": 2
   },
   "magnitude": {
    "present": false,
    "sentence_index": -1
   },
   "contribution": {
    "present": false,
    "sentence_index": -1
   }
  },
  "introduction": {
   "moves": [
    {
     "paragraph_index": 0,
     "move": "motivation"
    },
    {
     "paragraph_index": 0,
     "move": "other"
    },
    {
     "paragraph_index": 1,
     "move": "preview_of_approach"
    },
    {
     "paragraph_index": 2,
     "move": "related_literature_positioning"
    },
    {
     "paragraph_index": 2,
     "move": "literature_gap"
    },
    {
     "paragraph_index": 3,
     "move": "literature_gap"
    },
    {
     "paragraph_index": 3,
     "move": "contribution_statement"
    },
    {
     "paragraph_index": 3,
     "move": "preview_of_results"
    },
    {
     "paragraph_index": 4,
     "move": "roadmap"
    }
   ],
   "contribution_paragraph_index": 3,
   "contribution_types": [
    "new_evidence",
    "new_method",
    "new_data_or_measure"
   ],
   "roadmap_present": true,
   "literature_positioning_location": "inside_introduction"
  },
  "section_roles": [
   {
    "idx": 0,
    "role": "introduction"
   },
   {
    "idx": 1,
    "role": "empirical_design"
   },
   {
    "idx": 2,
    "role": "empirical_design"
   },
   {
    "idx": 3,
    "role": "results"
   },
   {
    "idx": 4,
    "role": "results"
   },
   {
    "idx": 5,
    "role": "conclusion"
   },
   {
    "idx": 6,
    "role": "appendix"
   }
  ],
  "results_presentation": {
   "economic_significance_discussed": true,
   "statistical_significance_language": "unclear",
   "magnitude_statements_in_intro": false
  },
  "robustness": {
   "location": "mixed",
   "types": [
    "subsample",
    "alternative_measure",
    "sensitivity_to_parameters"
   ]
  },
  "tables": {
   "convention_notes": "Notes open with 'This table presents...' and report year-by-year descriptive quantities (jump fractions, generalized correlations, Sharpe ratios). Sample sizes (N) and jump thresholds are given in the notes. Main comparison metric is generalized correlation between factor spaces. No regression coefficient tables with standard errors are visible.",
   "units_stated_in_notes": true,
   "standard_errors_description": "none"
  },
  "figures": {
   "uncertainty_shown": "unclear"
  },
  "conclusion": {
   "bounded_claims": true,
   "policy_implications": false,
   "open_questions": false,
   "restates_contribution": true
  },
  "hedging": {
   "level": "low",
   "typical_devices": [
    "approximately",
    "can be well approximated",
    "subtle",
    "roughly"
   ]
  },
  "claim_evidence": [
   {
    "claim": "Four stable continuous systematic factors and one stable jump factor explain individual stock comovement",
    "evidence_type": "figure",
    "evidence_ref": "Figures 1-2; Table II"
   },
   {
    "claim": "Continuous PCA factors are approximated by market, oil, finance and electricity portfolios",
    "evidence_type": "table",
    "evidence_ref": "Table III; Figures 3-4"
   },
   {
    "claim": "Factor loadings and explained variation vary over time while the loading span is stable",
    "evidence_type": "figure",
    "evidence_ref": "Figures 6-10; Table IV"
   },
   {
    "claim": "Fama-French-Carhart factors have time-varying span and explain less variation",
    "evidence_type": "figure",
    "evidence_ref": "Figures 9, 11; Table III"
   },
   {
    "claim": "Continuous factors earn an intraday risk premium that reverses overnight",
    "evidence_type": "table",
    "evidence_ref": "Table V; Figures 12-15"
   }
  ],
  "structural_summary": "The introduction poses three questions about systematic risk, then outlines a three-part approach: latent factors, high-frequency time variation, and a continuous/jump split. It positions this against constant-factor and characteristic-based studies and closes with a roadmap. Methodology and estimation sections set up a local factor model and PCA on high-frequency covariance matrices. Results cover the data, number of factors, balanced-panel checks, factor interpretation, time variation, and intraday versus overnight pricing. The conclusion enumerates the contributions.",
  "writing_profile_keywords": [
   "high-frequency econometrics",
   "latent factors",
   "PCA",
   "jump vs continuous",
   "time-varying loadings",
   "intraday vs overnight",
   "first-person singular",
   "figure-heavy"
  ],
  "confidence": 0.8,
  "notes": "The primary method is statistical factor analysis (PCA on high-frequency data), which has no exact schema category, so it is coded as 'other'. Section idx 4 appears to be a mis-parsed table heading that continues the results section. The estimation theory comes from the author's companion work (Pelger 2019)."
 },
 "deterministic_checks": {
  "roadmap_regex": true,
  "contribution_regex_in_intro": true,
  "magnitude_in_abstract": false,
  "abstract_n_sentences": 5,
  "intro_n_paragraphs": 5,
  "intro_n_words": 637,
  "causal_words_in_abstract": 0,
  "hedge_words_in_abstract": 0,
  "stars_mentioned": false,
  "clustered_se_mentioned": false,
  "internet_appendix_mentioned": true,
  "section_roles_rule": {
   "0": "introduction",
   "1": "empirical_design",
   "2": "empirical_design",
   "3": "results",
   "4": "other",
   "5": "conclusion",
   "6": "appendix"
  }
 },
 "schema_valid": true,
 "code_version": "87c73cd-dirty"
}