1{2 "validation_errors": [3 "Additional properties are not allowed ('typical_devices_note' was unexpected) @ hedging"4 ],5 "doc_id": "jof_12747",6 "ontology_version": "1.0",7 "prompt_version": "1.0",8 "model": "claude-opus-5-5",9 "effort": "medium",10 "annotated_at": "2026-11-03T23:06:07+00:00",11 "source_sha256": "1003c0785087714d103c43bb64b6d58afe608257d5da747efb905113a59af3fe",12 "parser": "pymupdf-jof-heuristics-1.3",13 "payload_chars": 35706,14 "usage": {15 "input": 11201,16 "output": 2497,17 "cache_read": 3968,18 "cache_write": 019 },20 "cost_usd": 0.095538,21 "request_id": "req_011CggqFZsej3JzRp4mSAUVi",22 "annotation": {23 "genre": "mixed_theory_empirical",24 "subfield": "asset_pricing",25 "subfield_secondary": "household_finance / mutual funds",26 "primary_method": "reduced_form_regression",27 "identification": {28 "estimand": "Sensitivity of aggregate monthly flows into active value and growth funds to option-implied investor preferences for downside protection and upside potential.",29 "strategy": "Time-series regressions of category-level active fund flows on pricing-kernel slopes and Prelec weighting parameters estimated from S&P 500 option prices, with sentiment, return and flow controls. Cross-sectional splits by fund tail properties, Active Share, and retirement versus retail funds support the hypothesized channel.",30 "key_assumptions": [31 "Option-implied pricing kernel reflects risk preferences of mutual fund investors via no-arbitrage",32 "Preference proxies are not merely proxies for investor sentiment",33 "Prelec functional form, or the moneyness range chosen for the PK slopes, captures the relevant tail preferences"34 ],35 "threats_addressed": [36 "investor sentiment confound (Baker-Wurgler control, low correlations)",37 "business cycle / recessions (NBER indicator)",38 "market volatility (VIX control)",39 "benchmark choice for active vs passive comparison",40 "differences in security characteristics held by active vs passive funds"41 ],42 "causal_language_strength": 3,43 "uses_causal_claim": true44 },45 "abstract_moves": {46 "question": {47 "present": false,48 "sentence_index": -149 },50 "gap": {51 "present": true,52 "sentence_index": 053 },54 "method": {55 "present": true,56 "sentence_index": 257 },58 "results": {59 "present": true,60 "sentence_index": 161 },62 "magnitude": {63 "present": false,64 "sentence_index": -165 },66 "contribution": {67 "present": true,68 "sentence_index": 269 }70 },71 "introduction": {72 "moves": [73 {74 "paragraph_index": 0,75 "move": "tension_or_puzzle"76 },77 {78 "paragraph_index": 0,79 "move": "preview_of_results"80 },81 {82 "paragraph_index": 1,83 "move": "preview_of_approach"84 },85 {86 "paragraph_index": 2,87 "move": "preview_of_results"88 },89 {90 "paragraph_index": 3,91 "move": "robustness_mention"92 },93 {94 "paragraph_index": 4,95 "move": "mechanism_or_theory_sketch"96 },97 {98 "paragraph_index": 5,99 "move": "preview_of_approach"100 },101 {102 "paragraph_index": 6,103 "move": "preview_of_results"104 },105 {106 "paragraph_index": 7,107 "move": "preview_of_approach"108 },109 {110 "paragraph_index": 8,111 "move": "robustness_mention"112 },113 {114 "paragraph_index": 9,115 "move": "identification_defense"116 },117 {118 "paragraph_index": 10,119 "move": "preview_of_results"120 },121 {122 "paragraph_index": 11,123 "move": "identification_defense"124 },125 {126 "paragraph_index": 12,127 "move": "preview_of_results"128 },129 {130 "paragraph_index": 13,131 "move": "preview_of_results"132 },133 {134 "paragraph_index": 14,135 "move": "preview_of_results"136 },137 {138 "paragraph_index": 15,139 "move": "related_literature_positioning"140 },141 {142 "paragraph_index": 16,143 "move": "contribution_statement"144 },145 {146 "paragraph_index": 17,147 "move": "policy_or_broader_implication"148 },149 {150 "paragraph_index": 18,151 "move": "roadmap"152 }153 ],154 "contribution_paragraph_index": 16,155 "contribution_types": [156 "new_evidence",157 "new_theory_or_model",158 "resolution_of_puzzle"159 ],160 "roadmap_present": true,161 "literature_positioning_location": "inside_introduction"162 },163 "section_roles": [164 {165 "idx": 0,166 "role": "introduction"167 },168 {169 "idx": 1,170 "role": "data"171 },172 {173 "idx": 2,174 "role": "results"175 },176 {177 "idx": 3,178 "role": "theory"179 },180 {181 "idx": 4,182 "role": "results"183 },184 {185 "idx": 5,186 "role": "results"187 },188 {189 "idx": 6,190 "role": "conclusion"191 }192 ],193 "results_presentation": {194 "economic_significance_discussed": true,195 "statistical_significance_language": "unclear",196 "magnitude_statements_in_intro": true197 },198 "robustness": {199 "location": "mixed",200 "types": [201 "alternative_measure",202 "subsample",203 "alternative_specification"204 ]205 },206 "tables": {207 "convention_notes": "Notes describe sample period, bootstrap design (100 funds, 400 repetitions), variable construction and controls. Summary statistics state units (millions USD, %). Regression tables are described as aggregate flow regressions with listed controls. Tercile sorts are defined in the notes. Standard error treatment is not visible in the excerpts.",208 "units_stated_in_notes": true,209 "standard_errors_description": "unclear"210 },211 "figures": {212 "uncertainty_shown": "unclear"213 },214 "conclusion": {215 "bounded_claims": false,216 "policy_implications": false,217 "open_questions": true,218 "restates_contribution": true219 },220 "hedging": {221 "level": "moderate",222 "typical_devices": [223 "'can help explain' / 'might help explain'",224 "'appear to' / 'appear consistent with'",225 "'to the extent that'",226 "explicit caveat that results need not imply investor segmentation",227 "'suggest' for implications"228 ],229 "typical_devices_note": ""230 },231 "claim_evidence": [232 {233 "claim": "Active value funds hedge downside risk better than passive benchmarks (lower co-skewness, downside beta, jump-risk hedging).",234 "evidence_type": "table",235 "evidence_ref": "Tables II-IV"236 },237 {238 "claim": "Active growth funds show higher skewness and positive idiosyncratic skewness relative to benchmarks.",239 "evidence_type": "table",240 "evidence_ref": "Tables II-III"241 },242 {243 "claim": "In an RDEU portfolio model, demand for active value (growth) funds rises with downside-protection (upside-seeking) preference.",244 "evidence_type": "figure",245 "evidence_ref": "Figures 2-3"246 },247 {248 "claim": "Investors would pay reasonable fees for access to active fund portfolios under the model.",249 "evidence_type": "figure",250 "evidence_ref": "Figure 4"251 },252 {253 "claim": "Option-implied downside (upside) preferences explain flows into active large value (growth) funds, with magnitudes comparable to past performance.",254 "evidence_type": "table",255 "evidence_ref": "Tables V-VI"256 },257 {258 "claim": "Flow sensitivity to preferences is stronger for funds with better tail properties and higher Active Share.",259 "evidence_type": "table",260 "evidence_ref": "Tables VIII-XI"261 },262 {263 "claim": "Retirement value funds are more sensitive to downside preferences; retail growth funds more sensitive to upside preferences.",264 "evidence_type": "table",265 "evidence_ref": "Table XII"266 }267 ],268 "structural_summary": "Opens with the puzzle of large active-fund assets despite mediocre performance. First documents distributional differences between active and passive funds using bootstrapped returns and several benchmarks. Then builds a rank-dependent expected utility portfolio model showing these differences matter to tail-sensitive investors. Next constructs option-implied preference proxies and runs time-series flow regressions with controls. Cross-sectional splits sharpen the channel. Closes by restating the findings and linking them to debates on managerial skill and non-EU preferences.",269 "writing_profile_keywords": [270 "puzzle-driven",271 "model-then-test",272 "bootstrap comparisons",273 "option-implied preferences",274 "cross-sectional validation",275 "alternative proxies",276 "literature complement"277 ],278 "confidence": 0.8,279 "notes": "Mixed paper: a numerical theory section sits inside the main text, followed by empirical tests. Standard error conventions and significance notation are not visible in the excerpts. The Internet Appendix holds robustness checks."280 },281 "deterministic_checks": {282 "roadmap_regex": true,283 "contribution_regex_in_intro": true,284 "magnitude_in_abstract": false,285 "abstract_n_sentences": 5,286 "intro_n_paragraphs": 19,287 "intro_n_words": 2771,288 "causal_words_in_abstract": 1,289 "hedge_words_in_abstract": 1,290 "stars_mentioned": false,291 "clustered_se_mentioned": false,292 "internet_appendix_mentioned": true,293 "section_roles_rule": {294 "0": "introduction",295 "1": "data",296 "2": "other",297 "3": "other",298 "4": "results",299 "5": "other",300 "6": "conclusion"301 }302 },303 "schema_valid": false,304 "code_version": "feb6c9e-dirty"305}