1{2 "validation_errors": [],3 "doc_id": "jof_12853",4 "ontology_version": "1.0",5 "prompt_version": "1.0",6 "model": "claude-opus-5-5",7 "effort": "medium",8 "annotated_at": "2026-11-03T23:12:19+00:00",9 "source_sha256": "c99829e37f39688f93175fcf4b39fc6f9ed99e07db2c2d6e86e91d6f82337ed6",10 "parser": "pymupdf-jof-heuristics-1.3",11 "payload_chars": 23368,12 "usage": {13 "input": 7811,14 "output": 2272,15 "cache_read": 3968,16 "cache_write": 017 },18 "cost_usd": 0.077478,19 "request_id": "req_011Cggqiw8qWKTKSQ6Ny3F78",20 "annotation": {21 "genre": "empirical",22 "subfield": "asset_pricing",23 "subfield_secondary": "mutual funds / asset management industry organization; textual analysis",24 "primary_method": "machine_learning_textual",25 "identification": {26 "estimand": "Associations between text-based fund uniqueness and fund starts by family size, fees, investor flows, returns, and flow-performance sensitivity",27 "strategy": "Text-based uniqueness built from prospectus strategy descriptions, used in pooled OLS regressions with fund controls and category-by-month fixed effects, compared against holdings- and returns-based uniqueness; comparison of start rates by family size",28 "key_assumptions": [29 "Prospectus text captures product differentiation that holdings and returns characteristics miss",30 "Controls such as expenses (which include marketing) absorb confounding drivers of flows"31 ],32 "threats_addressed": [33 "Measurement validity of the text measure, addressed with validation tests",34 "Overlap with holdings- and returns-based uniqueness",35 "Marketing expenses as an alternative channel for flows",36 "How fund starts are measured (counts vs. start rates scaled by family size)"37 ],38 "causal_language_strength": 2,39 "uses_causal_claim": false40 },41 "abstract_moves": {42 "question": {43 "present": true,44 "sentence_index": 045 },46 "gap": {47 "present": false,48 "sentence_index": -149 },50 "method": {51 "present": true,52 "sentence_index": 053 },54 "results": {55 "present": true,56 "sentence_index": 157 },58 "magnitude": {59 "present": false,60 "sentence_index": -161 },62 "contribution": {63 "present": false,64 "sentence_index": -165 }66 },67 "introduction": {68 "moves": [69 {70 "paragraph_index": 0,71 "move": "motivation"72 },73 {74 "paragraph_index": 0,75 "move": "literature_gap"76 },77 {78 "paragraph_index": 1,79 "move": "preview_of_approach"80 },81 {82 "paragraph_index": 1,83 "move": "mechanism_or_theory_sketch"84 },85 {86 "paragraph_index": 2,87 "move": "preview_of_results"88 },89 {90 "paragraph_index": 3,91 "move": "mechanism_or_theory_sketch"92 },93 {94 "paragraph_index": 4,95 "move": "preview_of_results"96 },97 {98 "paragraph_index": 5,99 "move": "preview_of_results"100 },101 {102 "paragraph_index": 5,103 "move": "mechanism_or_theory_sketch"104 },105 {106 "paragraph_index": 6,107 "move": "contribution_statement"108 },109 {110 "paragraph_index": 6,111 "move": "related_literature_positioning"112 },113 {114 "paragraph_index": 6,115 "move": "roadmap"116 }117 ],118 "contribution_paragraph_index": 6,119 "contribution_types": [120 "new_evidence",121 "new_data_or_measure"122 ],123 "roadmap_present": true,124 "literature_positioning_location": "inside_introduction"125 },126 "section_roles": [127 {128 "idx": 0,129 "role": "introduction"130 },131 {132 "idx": 1,133 "role": "theory"134 },135 {136 "idx": 2,137 "role": "data"138 },139 {140 "idx": 3,141 "role": "results"142 },143 {144 "idx": 4,145 "role": "results"146 },147 {148 "idx": 5,149 "role": "conclusion"150 },151 {152 "idx": 6,153 "role": "appendix"154 },155 {156 "idx": 7,157 "role": "appendix"158 },159 {160 "idx": 8,161 "role": "appendix"162 },163 {164 "idx": 9,165 "role": "appendix"166 }167 ],168 "results_presentation": {169 "economic_significance_discussed": true,170 "statistical_significance_language": "unclear",171 "magnitude_statements_in_intro": true172 },173 "robustness": {174 "location": "mixed",175 "types": [176 "alternative_measure",177 "sensitivity_to_parameters"178 ]179 },180 "tables": {181 "convention_notes": "Table notes describe the sample period (1999–2015/16), the estimator (pooled OLS, monthly fund data), the dependent variable, and fixed effects such as category-by-month. Panels split samples or dependent variables (e.g., gross vs. net returns). The visible note excerpts do not describe standard errors.",182 "units_stated_in_notes": true,183 "standard_errors_description": "unclear"184 },185 "figures": {186 "uncertainty_shown": "unclear"187 },188 "conclusion": {189 "bounded_claims": true,190 "policy_implications": false,191 "open_questions": false,192 "restates_contribution": true193 },194 "hedging": {195 "level": "moderate",196 "typical_devices": [197 "'one possible explanation' versus 'an alternative explanation'",198 "'associated with' wording",199 "'weak negative' qualifiers",200 "'we try to fill this gap'"201 ]202 },203 "claim_evidence": [204 {205 "claim": "Small and start-up fund families have higher start rates and launch more unique funds than large families",206 "evidence_type": "table",207 "evidence_ref": "Table IV; Figure 3"208 },209 {210 "claim": "New funds with above-median uniqueness attract about 50% (~$75M) more flows in their first 36 months",211 "evidence_type": "figure",212 "evidence_ref": "Figure 4; Table VII"213 },214 {215 "claim": "Text-based uniqueness predicts new-fund flows beyond holdings/returns uniqueness and expense controls",216 "evidence_type": "table",217 "evidence_ref": "Table VII"218 },219 {220 "claim": "Text uniqueness has only a weak negative relation with returns of established funds",221 "evidence_type": "table",222 "evidence_ref": "Table VIII"223 },224 {225 "claim": "Unique funds show lower flow-performance sensitivity",226 "evidence_type": "table",227 "evidence_ref": "Table IX"228 },229 {230 "claim": "The text-based uniqueness measure is valid, i.e., it identifies distinctive fund types",231 "evidence_type": "table",232 "evidence_ref": "Table III"233 },234 {235 "claim": "Smaller firms drive financial innovation, consistent with prior work",236 "evidence_type": "prior_literature",237 "evidence_ref": "Lerner (2006)"238 }239 ],240 "structural_summary": "The introduction opens with industry growth statistics, flags unexplored questions, and motivates text-based uniqueness with an illustrative fund example. It then previews results for new and established funds, sketches an equilibrium intuition based on Berk and Green, and closes with contributions, literature positioning, and a roadmap. A hypothesis/framework section precedes the data and validation section, which constructs the measure. Two results sections follow: fund starts, fees, and flows; then equilibrium returns and flow-performance sensitivity. A short conclusion restates the contributions. Appendices give example texts, the word-filtering procedure, cutoff sensitivity, and readability/tonality analysis.",241 "writing_profile_keywords": [242 "textual analysis",243 "new measure",244 "mutual funds",245 "product differentiation",246 "financial innovation",247 "validation tests",248 "pooled OLS",249 "hypothesis framework",250 "equilibrium intuition"251 ],252 "confidence": 0.8,253 "notes": "Section IV is labeled 'theory' by the role rule but is substantively a results section (equilibrium empirical results). Standard error treatment is not visible in the excerpts. Robustness evidence visible in the excerpts is mainly the information-ratio cutoff appendix and the comparison of alternative uniqueness measures."254 },255 "deterministic_checks": {256 "roadmap_regex": true,257 "contribution_regex_in_intro": false,258 "magnitude_in_abstract": false,259 "abstract_n_sentences": 5,260 "intro_n_paragraphs": 7,261 "intro_n_words": 1038,262 "causal_words_in_abstract": 0,263 "hedge_words_in_abstract": 0,264 "stars_mentioned": false,265 "clustered_se_mentioned": false,266 "internet_appendix_mentioned": false,267 "section_roles_rule": {268 "0": "introduction",269 "1": "theory",270 "2": "data",271 "3": "results",272 "4": "theory",273 "5": "conclusion",274 "6": "appendix",275 "7": "appendix",276 "8": "appendix",277 "9": "appendix"278 }279 },280 "schema_valid": true,281 "code_version": "87c73cd-dirty"282}