1{2 "validation_errors": [],3 "doc_id": "jof_12490",4 "ontology_version": "1.0",5 "prompt_version": "1.0",6 "model": "claude-opus-5-5",7 "effort": "medium",8 "annotated_at": "2026-11-03T22:56:40+00:00",9 "source_sha256": "22332f923950a346e53845513bb3c898a690075826c1ff3f9c38cbd237058120",10 "parser": "pymupdf-jof-heuristics-1.3",11 "payload_chars": 24514,12 "usage": {13 "input": 8499,14 "output": 2273,15 "cache_read": 3968,16 "cache_write": 017 },18 "cost_usd": 0.08025,19 "request_id": "req_011CggpXrUs8VJiNd5tJs5TJ",20 "annotation": {21 "genre": "mixed_theory_empirical",22 "subfield": "corporate_finance",23 "subfield_secondary": "corporate cash holdings / governance",24 "primary_method": "instrumental_variables",25 "identification": {26 "estimand": "Effect of financial portfolio size on the share of the portfolio held in risky assets; marginal value of risky versus safe financial assets",27 "strategy": "Panel regressions run by OLS and by 2SLS. Portfolio size is instrumented with unexpected operating cash flow innovations. Value tests regress stock returns and market-to-book on financial asset holdings, following the Faulkender–Wang approach.",28 "key_assumptions": [29 "Unexpected cash flow innovations shift portfolio size but affect portfolio composition only through size",30 "Fair value footnote disclosures accurately capture asset classes"31 ],32 "threats_addressed": [33 "Joint determination of portfolio size and composition (endogeneity)"34 ],35 "causal_language_strength": 3,36 "uses_causal_claim": true37 },38 "abstract_moves": {39 "question": {40 "present": true,41 "sentence_index": 242 },43 "gap": {44 "present": false,45 "sentence_index": -146 },47 "method": {48 "present": true,49 "sentence_index": 250 },51 "results": {52 "present": true,53 "sentence_index": 354 },55 "magnitude": {56 "present": true,57 "sentence_index": 158 },59 "contribution": {60 "present": true,61 "sentence_index": 462 }63 },64 "introduction": {65 "moves": [66 {67 "paragraph_index": 0,68 "move": "motivation"69 },70 {71 "paragraph_index": 0,72 "move": "tension_or_puzzle"73 },74 {75 "paragraph_index": 0,76 "move": "preview_of_approach"77 },78 {79 "paragraph_index": 1,80 "move": "preview_of_approach"81 },82 {83 "paragraph_index": 1,84 "move": "mechanism_or_theory_sketch"85 },86 {87 "paragraph_index": 2,88 "move": "mechanism_or_theory_sketch"89 },90 {91 "paragraph_index": 3,92 "move": "preview_of_results"93 },94 {95 "paragraph_index": 4,96 "move": "identification_defense"97 },98 {99 "paragraph_index": 4,100 "move": "preview_of_results"101 },102 {103 "paragraph_index": 5,104 "move": "preview_of_results"105 },106 {107 "paragraph_index": 5,108 "move": "robustness_mention"109 },110 {111 "paragraph_index": 6,112 "move": "contribution_statement"113 },114 {115 "paragraph_index": 6,116 "move": "related_literature_positioning"117 },118 {119 "paragraph_index": 7,120 "move": "policy_or_broader_implication"121 }122 ],123 "contribution_paragraph_index": 6,124 "contribution_types": [125 "new_evidence",126 "new_data_or_measure",127 "new_theory_or_model"128 ],129 "roadmap_present": false,130 "literature_positioning_location": "both"131 },132 "section_roles": [133 {134 "idx": 0,135 "role": "introduction"136 },137 {138 "idx": 1,139 "role": "data"140 },141 {142 "idx": 2,143 "role": "results"144 },145 {146 "idx": 3,147 "role": "theory"148 },149 {150 "idx": 4,151 "role": "results"152 },153 {154 "idx": 5,155 "role": "results"156 },157 {158 "idx": 6,159 "role": "conclusion"160 },161 {162 "idx": 7,163 "role": "appendix"164 },165 {166 "idx": 8,167 "role": "appendix"168 },169 {170 "idx": 9,171 "role": "appendix"172 },173 {174 "idx": 10,175 "role": "appendix"176 }177 ],178 "results_presentation": {179 "economic_significance_discussed": true,180 "statistical_significance_language": "unclear",181 "magnitude_statements_in_intro": true182 },183 "robustness": {184 "location": "within_results",185 "types": [186 "alternative_measure",187 "alternative_specification"188 ]189 },190 "tables": {191 "convention_notes": "Table notes define variables and write out the regression equations, including the 2SLS first stage. OLS and 2SLS columns are reported side by side. Descriptive tables give aggregate value-weighted figures, with equal-weighted versions in an appendix. Appendix tables reproduce firm disclosures from Google, Apple and Intel as illustrations.",192 "units_stated_in_notes": true,193 "standard_errors_description": "unclear"194 },195 "figures": {196 "uncertainty_shown": "no"197 },198 "conclusion": {199 "bounded_claims": true,200 "policy_implications": true,201 "open_questions": true,202 "restates_contribution": true203 },204 "hedging": {205 "level": "moderate",206 "typical_devices": [207 "'consistent with' framing of results against model predictions",208 "'suggest' for 2SLS estimates",209 "explicit sample limitation acknowledged in conclusion"210 ]211 },212 "claim_evidence": [213 {214 "claim": "Risky assets make up about 40% of firms' financial portfolios and about 6% of book assets.",215 "evidence_type": "table",216 "evidence_ref": "Table I"217 },218 {219 "claim": "The share held in risky assets rises from about 10% to 40% across portfolio-size quintiles, and allocation variation is higher in the top quintiles.",220 "evidence_type": "table",221 "evidence_ref": "Table II"222 },223 {224 "claim": "A one-point increase in portfolio size raises the risky share by about 30 bp (2SLS estimate).",225 "evidence_type": "table",226 "evidence_ref": "Table III"227 },228 {229 "claim": "Permanently reinvested earnings relate positively to total financial assets but insignificantly to risky assets.",230 "evidence_type": "table",231 "evidence_ref": "Table IV"232 },233 {234 "claim": "Poorly governed firms hold more risky assets.",235 "evidence_type": "table",236 "evidence_ref": "Table V"237 },238 {239 "claim": "Risky financial assets are valued at a 13–22% discount relative to safe assets.",240 "evidence_type": "table",241 "evidence_ref": "Table VII"242 },243 {244 "claim": "Constrained firms optimally avoid risky assets; unconstrained firms are at best indifferent to them.",245 "evidence_type": "equation_or_proof",246 "evidence_ref": "Section III; Internet Appendix"247 },248 {249 "claim": "There is no uniform tax incentive to invest in risky financial assets.",250 "evidence_type": "equation_or_proof",251 "evidence_ref": "Appendix C"252 }253 ],254 "structural_summary": "The paper opens with press anecdotes that challenge the assumption that corporate 'cash' is in fact cash. It introduces hand-collected fair value data made available by SFAS 157, then describes portfolio composition. A theory section sketches a precautionary-savings model extended to risky and illiquid assets, with the formal derivation in an internet appendix. Empirical tests follow: univariate sorts, then OLS/2SLS determinants (taxes, agency, CEO traits), then valuation and financing-source tests. The conclusion frames the findings as a largely unregulated shadow asset management industry.",255 "writing_profile_keywords": [256 "hand-collected data",257 "model-consistent predictions",258 "precautionary savings",259 "2SLS",260 "shadow hedge fund framing",261 "policy implications",262 "anecdotal motivation"263 ],264 "confidence": 0.8,265 "notes": "Theory is summarized in the main text and formally derived in an internet appendix. Standard error treatment is not visible in the excerpts. Section II is descriptive results."266 },267 "deterministic_checks": {268 "roadmap_regex": false,269 "contribution_regex_in_intro": false,270 "magnitude_in_abstract": true,271 "abstract_n_sentences": 6,272 "intro_n_paragraphs": 8,273 "intro_n_words": 1027,274 "causal_words_in_abstract": 0,275 "hedge_words_in_abstract": 1,276 "stars_mentioned": false,277 "clustered_se_mentioned": false,278 "internet_appendix_mentioned": true,279 "section_roles_rule": {280 "0": "introduction",281 "1": "data",282 "2": "other",283 "3": "theory",284 "4": "other",285 "5": "other",286 "6": "conclusion",287 "7": "appendix",288 "8": "appendix",289 "9": "appendix",290 "10": "appendix"291 }292 },293 "schema_valid": true,294 "code_version": "feb6c9e-dirty"295}