1{2 "validation_errors": [],3 "doc_id": "jof_12704",4 "ontology_version": "1.0",5 "prompt_version": "1.0",6 "model": "claude-opus-5-5",7 "effort": "medium",8 "annotated_at": "2026-11-03T23:03:25+00:00",9 "source_sha256": "1b50702d7ad6b1fce3a2f80dea73ad83ed1304ce5fd912c1f8f01d409f1c241d",10 "parser": "pymupdf-jof-heuristics-1.3",11 "payload_chars": 25609,12 "usage": {13 "input": 8535,14 "output": 2236,15 "cache_read": 3968,16 "cache_write": 017 },18 "cost_usd": 0.079654,19 "request_id": "req_011Cggq3fD3bwhUtKUdbbWbB",20 "annotation": {21 "genre": "empirical",22 "subfield": "corporate_finance",23 "subfield_secondary": "behavioral_finance",24 "primary_method": "reduced_form_regression",25 "identification": {26 "estimand": "Cross-city dispersion in financial misconduct rates beyond sampling variation, and the association of city FM rates with measures of local social norms (non-FM misbehavior).",27 "strategy": "Negative binomial and Poisson count regressions at the city and city-year level testing excess variance and city fixed effects; cross-sectional regressions of FM on a misbehavior index; controls for regional characteristics, enforcement proxies, and firm characteristics; peer logit regressions.",28 "key_assumptions": [29 "Non-FM misbehavior measures proxy for city-level ethical norms",30 "Detected FM reflects committed FM after controlling for enforcement proxies",31 "Observable controls capture relevant firm and regional heterogeneity"32 ],33 "threats_addressed": [34 "sampling variation / chance clustering",35 "regional enforcement differences (detection and deterrence)",36 "firm heterogeneity in benefits of FM",37 "regional demographic and economic characteristics",38 "reflection problem (acknowledged caveat)"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": false,54 "sentence_index": -155 },56 "results": {57 "present": true,58 "sentence_index": 159 },60 "magnitude": {61 "present": true,62 "sentence_index": 063 },64 "contribution": {65 "present": false,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": "tension_or_puzzle"78 },79 {80 "paragraph_index": 1,81 "move": "preview_of_approach"82 },83 {84 "paragraph_index": 2,85 "move": "preview_of_results"86 },87 {88 "paragraph_index": 4,89 "move": "preview_of_results"90 },91 {92 "paragraph_index": 8,93 "move": "preview_of_results"94 },95 {96 "paragraph_index": 10,97 "move": "roadmap"98 },99 {100 "paragraph_index": 10,101 "move": "preview_of_approach"102 },103 {104 "paragraph_index": 11,105 "move": "preview_of_approach"106 },107 {108 "paragraph_index": 12,109 "move": "preview_of_results"110 },111 {112 "paragraph_index": 13,113 "move": "preview_of_results"114 },115 {116 "paragraph_index": 14,117 "move": "identification_defense"118 },119 {120 "paragraph_index": 14,121 "move": "mechanism_or_theory_sketch"122 },123 {124 "paragraph_index": 15,125 "move": "related_literature_positioning"126 },127 {128 "paragraph_index": 15,129 "move": "policy_or_broader_implication"130 },131 {132 "paragraph_index": 16,133 "move": "related_literature_positioning"134 },135 {136 "paragraph_index": 16,137 "move": "policy_or_broader_implication"138 }139 ],140 "contribution_paragraph_index": 15,141 "contribution_types": [142 "new_evidence",143 "new_data_or_measure"144 ],145 "roadmap_present": true,146 "literature_positioning_location": "inside_introduction"147 },148 "section_roles": [149 {150 "idx": 0,151 "role": "introduction"152 },153 {154 "idx": 1,155 "role": "theory"156 },157 {158 "idx": 2,159 "role": "data"160 },161 {162 "idx": 3,163 "role": "results"164 },165 {166 "idx": 4,167 "role": "results"168 },169 {170 "idx": 5,171 "role": "robustness"172 },173 {174 "idx": 6,175 "role": "conclusion"176 },177 {178 "idx": 7,179 "role": "appendix"180 }181 ],182 "results_presentation": {183 "economic_significance_discussed": true,184 "statistical_significance_language": "mixed",185 "magnitude_statements_in_intro": true186 },187 "robustness": {188 "location": "mixed",189 "types": [190 "alternative_specification",191 "subsample",192 "alternative_measure",193 "alternative_sample_period"194 ]195 },196 "tables": {197 "convention_notes": "Table notes open by naming the estimator (negative binomial, Poisson, panel logit) and defining the dependent variable (area-level count of FM firm-years). Appendix tables repeat the regressions with alternative constructions of the misbehavior index, area-year units, and a large-firm subsample.",198 "units_stated_in_notes": true,199 "standard_errors_description": "unclear"200 },201 "figures": {202 "uncertainty_shown": "unclear"203 },204 "conclusion": {205 "bounded_claims": true,206 "policy_implications": false,207 "open_questions": true,208 "restates_contribution": true209 },210 "hedging": {211 "level": "moderate",212 "typical_devices": [213 "appear to",214 "suggest",215 "cannot definitively rule out",216 "remain agnostic about ultimate causes",217 "explicit caveat on reflection problem"218 ]219 },220 "claim_evidence": [221 {222 "claim": "FM rates cluster in some cities more than chance predicts, with roughly threefold differences across cities.",223 "evidence_type": "table",224 "evidence_ref": "Table I, Table II"225 },226 {227 "claim": "City fixed effects remain significant in panel regressions at the city-year level.",228 "evidence_type": "table",229 "evidence_ref": "Table III"230 },231 {232 "claim": "Each of five non-FM misbehavior measures is positively related to city FM rates; the combined index has a t-statistic above four.",233 "evidence_type": "table",234 "evidence_ref": "Table IV"235 },236 {237 "claim": "City FM rates and the misbehavior index ranking are strongly correlated (rho of about 0.60).",238 "evidence_type": "figure",239 "evidence_ref": "Figure 2"240 },241 {242 "claim": "Including the misbehavior index eliminates significant excess clustering of FM across cities.",243 "evidence_type": "table",244 "evidence_ref": "Table IV"245 },246 {247 "claim": "Enforcement proxies and firm characteristics do not explain the city differences in FM.",248 "evidence_type": "table",249 "evidence_ref": "Table VII, Table VIII"250 }251 ],252 "structural_summary": "Opens with the gap between Becker-style crime models and observed spatial variation in crime, then previews numbered findings and walks through the sections. A stylized cost-benefit model motivates three candidate explanations: norms, enforcement, and firm heterogeneity. The analysis first measures excess clustering of FM with count models, then relates FM to observable regional traits and to an index of non-FM misbehavior. Separate sections rule out enforcement and firm-characteristic explanations. The paper closes by acknowledging that the sources of norms remain unidentified and flags them for future research.",253 "writing_profile_keywords": [254 "social norms",255 "geography",256 "count models",257 "measurement-first",258 "alternative explanations",259 "rare events",260 "culture and finance"261 ],262 "confidence": 0.82,263 "notes": "The introduction's numbered findings list is split across paragraphs 2-9. Causal language is cautious: norms 'appear' to matter, and the authors explicitly acknowledge the reflection problem and that the sources of norms are unidentified."264 },265 "deterministic_checks": {266 "roadmap_regex": false,267 "contribution_regex_in_intro": true,268 "magnitude_in_abstract": false,269 "abstract_n_sentences": 4,270 "intro_n_paragraphs": 17,271 "intro_n_words": 1216,272 "causal_words_in_abstract": 0,273 "hedge_words_in_abstract": 1,274 "stars_mentioned": false,275 "clustered_se_mentioned": false,276 "internet_appendix_mentioned": false,277 "section_roles_rule": {278 "0": "introduction",279 "1": "theory",280 "2": "data",281 "3": "other",282 "4": "other",283 "5": "robustness",284 "6": "conclusion",285 "7": "appendix"286 }287 },288 "schema_valid": true,289 "code_version": "feb6c9e-dirty"290}