1{2 "validation_errors": [],3 "doc_id": "jof_12514",4 "ontology_version": "1.0",5 "prompt_version": "1.0",6 "model": "claude-opus-5-5",7 "effort": "medium",8 "annotated_at": "2026-11-03T22:57:52+00:00",9 "source_sha256": "7dd8bfbef11797e52089a25c38b62f4ce1788c8e2a4450f2c2af25eda6b22d01",10 "parser": "pymupdf-jof-heuristics-1.3",11 "payload_chars": 18515,12 "usage": {13 "input": 6266,14 "output": 2147,15 "cache_read": 3968,16 "cache_write": 017 },18 "cost_usd": 0.068798,19 "request_id": "req_011Cggpd87721PchFpgzdAHV",20 "annotation": {21 "genre": "empirical",22 "subfield": "household_finance",23 "subfield_secondary": "financial_intermediation_banking",24 "primary_method": "reduced_form_regression",25 "identification": {26 "estimand": "The share of cross-sectional variation in client portfolio risky share and home bias that is attributable to advisor-specific effects, compared with the share explained by investor attributes.",27 "strategy": "Panel regressions with advisor fixed effects and year fixed effects. A two-way fixed effects design adds investor fixed effects and uses clients forced to switch advisors after the advisor's death, retirement or resignation, then tracks convergence toward the new advisor's allocation. Advisor fixed effects are then regressed on advisor attributes, including the advisor's own portfolio.",28 "key_assumptions": [29 "Forced switches are unrelated to changes in client preferences.",30 "Investor fixed effects absorb persistent unobserved client preferences.",31 "Investors are not matched to advisors on unobserved characteristics, or such matching is controlled for by the switcher design."32 ],33 "threats_addressed": [34 "Matching between investors and advisors on omitted characteristics",35 "Unobserved client preferences",36 "Self-directed investments initiated by the client",37 "Client-initiated switches that coincide with preference changes"38 ],39 "causal_language_strength": 3,40 "uses_causal_claim": true41 },42 "abstract_moves": {43 "question": {44 "present": false,45 "sentence_index": -146 },47 "gap": {48 "present": false,49 "sentence_index": -150 },51 "method": {52 "present": true,53 "sentence_index": 054 },55 "results": {56 "present": true,57 "sentence_index": 158 },59 "magnitude": {60 "present": true,61 "sentence_index": 462 },63 "contribution": {64 "present": false,65 "sentence_index": -166 }67 },68 "introduction": {69 "moves": [70 {71 "paragraph_index": 0,72 "move": "motivation"73 },74 {75 "paragraph_index": 0,76 "move": "literature_gap"77 },78 {79 "paragraph_index": 0,80 "move": "tension_or_puzzle"81 },82 {83 "paragraph_index": 0,84 "move": "preview_of_approach"85 },86 {87 "paragraph_index": 1,88 "move": "preview_of_results"89 },90 {91 "paragraph_index": 2,92 "move": "preview_of_results"93 },94 {95 "paragraph_index": 2,96 "move": "other"97 },98 {99 "paragraph_index": 3,100 "move": "identification_defense"101 },102 {103 "paragraph_index": 3,104 "move": "preview_of_results"105 },106 {107 "paragraph_index": 4,108 "move": "contribution_statement"109 },110 {111 "paragraph_index": 5,112 "move": "roadmap"113 }114 ],115 "contribution_paragraph_index": 4,116 "contribution_types": [117 "new_evidence",118 "new_data_or_measure"119 ],120 "roadmap_present": true,121 "literature_positioning_location": "inside_introduction"122 },123 "section_roles": [124 {125 "idx": 0,126 "role": "introduction"127 },128 {129 "idx": 1,130 "role": "data"131 },132 {133 "idx": 2,134 "role": "results"135 },136 {137 "idx": 3,138 "role": "results"139 },140 {141 "idx": 4,142 "role": "results"143 },144 {145 "idx": 5,146 "role": "conclusion"147 }148 ],149 "results_presentation": {150 "economic_significance_discussed": true,151 "statistical_significance_language": "unclear",152 "magnitude_statements_in_intro": true153 },154 "robustness": {155 "location": "within_results",156 "types": [157 "other"158 ]159 },160 "tables": {161 "convention_notes": "Notes define the dependent variables (risky share, home bias), the sample period and the fixed effects included. Tables are organized in panels by outcome (risky share vs. home bias). Descriptive statistics are reported as percentiles.",162 "units_stated_in_notes": true,163 "standard_errors_description": "unclear"164 },165 "figures": {166 "uncertainty_shown": "mixed"167 },168 "conclusion": {169 "bounded_claims": false,170 "policy_implications": false,171 "open_questions": false,172 "restates_contribution": true173 },174 "hedging": {175 "level": "moderate",176 "typical_devices": [177 "Offering alternative interpretations of the advisor fixed effects",178 "An explicit caveats-and-robustness subsection",179 "Framing the matching concern as a hypothesis that is then tested"180 ]181 },182 "claim_evidence": [183 {184 "claim": "Investor characteristics explain only about 12% of the variation in risky share.",185 "evidence_type": "table",186 "evidence_ref": "Table II"187 },188 {189 "claim": "Advisor fixed effects raise adjusted R2 from 12% to 30% for risky share, and add 24% for home bias.",190 "evidence_type": "table",191 "evidence_ref": "Table II"192 },193 {194 "claim": "Moving from the 25th to the 75th percentile of advisor effects shifts risky share by 20 pp and home bias by 32 pp.",195 "evidence_type": "figure",196 "evidence_ref": "Figure 3"197 },198 {199 "claim": "Advisor effects remain important after controlling for investor fixed effects among forced switchers.",200 "evidence_type": "table",201 "evidence_ref": "Table III"202 },203 {204 "claim": "An advisor's own risky share is the strongest predictor of the advisor fixed effect.",205 "evidence_type": "table",206 "evidence_ref": "Table IV; Figure 5"207 },208 {209 "claim": "Advised portfolios cost about 2.5% per year, roughly 1.5% more than life cycle funds.",210 "evidence_type": "figure",211 "evidence_ref": "Figure 6; Section IV"212 },213 {214 "claim": "Advisors' gross and net alphas are estimated against passive factor benchmarks.",215 "evidence_type": "table",216 "evidence_ref": "Table V"217 }218 ],219 "structural_summary": "The introduction motivates the topic with how widely households use advisors, then sets the competing views of underperformance and \"interior decoration\" against each other. It previews variance-decomposition results with R2 magnitudes and raises the matching interpretation. It addresses that interpretation with a forced-switcher two-way fixed effects design, then states three contributions and gives a roadmap. The body proceeds through the data, customization regressions (with caveats, investor fixed effects, and advisor attributes), performance alphas, and a fee decomposition. The conclusion restates three findings with magnitudes.",220 "writing_profile_keywords": [221 "administrative data",222 "variance decomposition",223 "advisor fixed effects",224 "two-way fixed effects",225 "forced switchers",226 "adjusted R2 magnitudes",227 "fee decomposition",228 "one-size-fits-all"229 ],230 "confidence": 0.8,231 "notes": "The conclusion excerpt is truncated, so policy implications may appear later in it but are not visible. The standard error type is not visible in the table note excerpts. Some figures show 95% confidence intervals (Figures 1 and 4), while others show distributions or averages."232 },233 "deterministic_checks": {234 "roadmap_regex": true,235 "contribution_regex_in_intro": true,236 "magnitude_in_abstract": true,237 "abstract_n_sentences": 5,238 "intro_n_paragraphs": 6,239 "intro_n_words": 797,240 "causal_words_in_abstract": 0,241 "hedge_words_in_abstract": 0,242 "stars_mentioned": false,243 "clustered_se_mentioned": false,244 "internet_appendix_mentioned": false,245 "section_roles_rule": {246 "0": "introduction",247 "1": "other",248 "2": "other",249 "3": "other",250 "4": "other",251 "5": "conclusion"252 }253 },254 "schema_valid": true,255 "code_version": "feb6c9e-dirty"256}