1{2 "validation_errors": [],3 "doc_id": "jof_12699",4 "ontology_version": "1.0",5 "prompt_version": "1.0",6 "model": "claude-opus-5-5",7 "effort": "medium",8 "annotated_at": "2026-11-03T23:03:43+00:00",9 "source_sha256": "7602100a6164a6e70057c842cb89792ea14c67804412c027cc44a44acb0d8c0e",10 "parser": "pymupdf-jof-heuristics-1.3",11 "payload_chars": 24126,12 "usage": {13 "input": 7815,14 "output": 2165,15 "cache_read": 3968,16 "cache_write": 017 },18 "cost_usd": 0.075354,19 "request_id": "req_011Cggq54qsG6DMdXyTrK64P",20 "annotation": {21 "genre": "empirical",22 "subfield": "asset_pricing",23 "subfield_secondary": "mutual fund performance / institutional herding",24 "primary_method": "reduced_form_regression",25 "identification": {26 "estimand": "Predictive association between fund-level herding tendency and future fund performance (alphas, returns), with skill as the posited driver",27 "strategy": "Construct a fund-level herding measure from cross-sectional regressions of fund trades on past institutional trades; run portfolio sorts and predictive panel and Fama-MacBeth regressions; add auxiliary tests (stocks with light institutional trading, time-varying opportunities, persistence, anticipating crowd trades, career concerns)",28 "key_assumptions": [29 "The herding measure filters out common styles, momentum trading and common information signals",30 "Predictability reflects skill rather than chance or price pressure from institutional herds",31 "Risk adjustment through factor models is adequate"32 ],33 "threats_addressed": [34 "price pressure from institutional herds (test on stocks not heavily traded)",35 "chance (performance persistence)",36 "overlap with past performance and other skill measures (regression controls)",37 "stock characteristics and serial dependence in a fund's own trades (alternative herding measures)",38 "choice of performance measure (holdings-based DGTW CS and Grinblatt-Titman)"39 ],40 "causal_language_strength": 3,41 "uses_causal_claim": true42 },43 "abstract_moves": {44 "question": {45 "present": false,46 "sentence_index": -147 },48 "gap": {49 "present": false,50 "sentence_index": -151 },52 "method": {53 "present": true,54 "sentence_index": 155 },56 "results": {57 "present": true,58 "sentence_index": 059 },60 "magnitude": {61 "present": true,62 "sentence_index": 263 },64 "contribution": {65 "present": true,66 "sentence_index": 167 }68 },69 "introduction": {70 "moves": [71 {72 "paragraph_index": 0,73 "move": "motivation"74 },75 {76 "paragraph_index": 0,77 "move": "literature_gap"78 },79 {80 "paragraph_index": 1,81 "move": "preview_of_approach"82 },83 {84 "paragraph_index": 2,85 "move": "mechanism_or_theory_sketch"86 },87 {88 "paragraph_index": 3,89 "move": "preview_of_results"90 },91 {92 "paragraph_index": 4,93 "move": "preview_of_results"94 },95 {96 "paragraph_index": 5,97 "move": "contribution_statement"98 },99 {100 "paragraph_index": 5,101 "move": "related_literature_positioning"102 },103 {104 "paragraph_index": 6,105 "move": "contribution_statement"106 },107 {108 "paragraph_index": 6,109 "move": "related_literature_positioning"110 },111 {112 "paragraph_index": 7,113 "move": "roadmap"114 }115 ],116 "contribution_paragraph_index": 5,117 "contribution_types": [118 "new_evidence",119 "new_data_or_measure"120 ],121 "roadmap_present": true,122 "literature_positioning_location": "inside_introduction"123 },124 "section_roles": [125 {126 "idx": 0,127 "role": "introduction"128 },129 {130 "idx": 1,131 "role": "data"132 },133 {134 "idx": 2,135 "role": "results"136 },137 {138 "idx": 3,139 "role": "results"140 },141 {142 "idx": 4,143 "role": "results"144 },145 {146 "idx": 5,147 "role": "robustness"148 },149 {150 "idx": 6,151 "role": "results"152 },153 {154 "idx": 7,155 "role": "conclusion"156 }157 ],158 "results_presentation": {159 "economic_significance_discussed": true,160 "statistical_significance_language": "unclear",161 "magnitude_statements_in_intro": true162 },163 "robustness": {164 "location": "main_text_section",165 "types": [166 "alternative_measure",167 "alternative_specification"168 ]169 },170 "tables": {171 "convention_notes": "Table notes describe variable construction in detail, often referring back to Table I. Results use decile and quintile portfolio sorts, Fama-MacBeth and panel predictive regressions, and a logit regression. Returns are mostly monthly, with alphas from factor models such as Carhart.",172 "units_stated_in_notes": true,173 "standard_errors_description": "unclear"174 },175 "figures": {176 "uncertainty_shown": "no_figures"177 },178 "conclusion": {179 "bounded_claims": false,180 "policy_implications": false,181 "open_questions": false,182 "restates_contribution": true183 },184 "hedging": {185 "level": "moderate",186 "typical_devices": [187 "suggests",188 "may/might in theory discussion",189 "is related to",190 "in line with theory"191 ]192 },193 "claim_evidence": [194 {195 "claim": "The highest-herding decile of funds underperforms the antiherding decile by about 2.28% per year, and by 1.92% on a four-factor alpha basis",196 "evidence_type": "table",197 "evidence_ref": "Table II"198 },199 {200 "claim": "Herding predicts performance beyond past performance, other fund characteristics and other skill measures",201 "evidence_type": "table",202 "evidence_ref": "Table IV"203 },204 {205 "claim": "Among stocks not heavily traded by institutions, antiherding funds' large bets outperform herding funds' holdings by about 38 bps per month in Carhart alpha",206 "evidence_type": "table",207 "evidence_ref": "Table V"208 },209 {210 "claim": "The herding-antiherding performance gap widens when investment opportunities are greater",211 "evidence_type": "table",212 "evidence_ref": "Table VI"213 },214 {215 "claim": "The performance gap persists for up to two years",216 "evidence_type": "table",217 "evidence_ref": "Table VII"218 },219 {220 "claim": "Trades of antiherding funds predict future institutional trades",221 "evidence_type": "table",222 "evidence_ref": "Table VIII"223 },224 {225 "claim": "Herding reveals skill mainly among less experienced managers",226 "evidence_type": "table",227 "evidence_ref": "Tables IX-XI"228 },229 {230 "claim": "Fund flows show an insignificant negative relation to herding",231 "evidence_type": "table",232 "evidence_ref": "Table XV"233 }234 ],235 "structural_summary": "The paper motivates the question with theoretical herding models that link ability to deviation from the crowd, and notes the lack of empirical evidence. It then builds a new dynamic fund-level herding measure and shows that it negatively predicts fund returns, using portfolio sorts and predictive regressions. Four sets of tests attribute the predictability to skill. A further section ties herding to career concerns and managerial experience. Robustness checks use alternative herding and performance measures, followed by a short analysis of investor flows and a summary conclusion.",236 "writing_profile_keywords": [237 "new measure",238 "predictability",239 "skill attribution",240 "theory-motivated tests",241 "portfolio sorts",242 "mutual funds",243 "multi-pronged evidence"244 ],245 "confidence": 0.82,246 "notes": "Abstract sentence 4 bundles the mechanism tests; the abstract text also runs into the opening of the introduction. Standard error conventions and significance notation are not visible in the excerpts."247 },248 "deterministic_checks": {249 "roadmap_regex": true,250 "contribution_regex_in_intro": true,251 "magnitude_in_abstract": true,252 "abstract_n_sentences": 5,253 "intro_n_paragraphs": 8,254 "intro_n_words": 1205,255 "causal_words_in_abstract": 0,256 "hedge_words_in_abstract": 1,257 "stars_mentioned": false,258 "clustered_se_mentioned": false,259 "internet_appendix_mentioned": false,260 "section_roles_rule": {261 "0": "introduction",262 "1": "other",263 "2": "other",264 "3": "other",265 "4": "other",266 "5": "robustness",267 "6": "other",268 "7": "conclusion"269 }270 },271 "schema_valid": true,272 "code_version": "feb6c9e-dirty"273}