Replication Guide - 65 Paper Corpus

1. Overview and Key Innovation

This guide documents the complete methodology for constructing a corpus of 65 peer-reviewed papers on mutual fund style drift. The corpus was built through a two-stage process: (1) parallel independent searches using different strategies, and (2) unified keyword-based relevance classification.

Key Innovation

Rather than relying solely on journal prestige or citation counts, we applied domain-specific keyword filtering that mirrors how experts identify relevant literature. Papers must contain established terminology from the style drift literature (e.g., "style drift", "active share", "closet indexing") to be classified as relevant.

Phase 1: 58-Corpus          Phase 2: 53-Corpus          Phase 3: Merge
     |                            |                          |
     v                            v                          v
+-------------+             +-------------+            +-------------+
| 19 OpenAlex |             | Broad Search|            | Load Both   |
|   Queries   |             | Top Journals|            |   Corpora   |
+-------------+             +-------------+            +-------------+
     |                            |                          |
     v                            v                          v
+-------------+             +-------------+            +-------------+
| 5400+ Raw   |             | 2600+ Raw   |            | Deduplicate |
|   Papers    |             |   Papers    |            |   by DOI    |
+-------------+             +-------------+            +-------------+
     |                            |                          |
     v                            v                          v
+-------------+             +-------------+            +-------------+
| Stage 1-2   |             | Journal +   |            | 111 Unique  |
| Filters     |             | Citation    |            |   Papers    |
+-------------+             +-------------+            +-------------+
     |                            |                          |
     v                            v                          v
+-------------+             +-------------+            +-------------+
| Stage 3-4   |             | Weak Kw     |            | Keyword     |
| Journals+Kw |             | Filter      |            | Classify    |
+-------------+             +-------------+            +-------------+
     |                            |                          |
     v                            v                          v
+-------------+             +-------------+            +-------------+
|  58 Papers  |             |  53 Papers  |            |  65 RELEVANT|
|   (100%)    |             |  (13.2%)    |            |  46 REVIEW  |
+-------------+             +-------------+            +-------------+

2. Prerequisites and Dependencies

Python Environment

Requirement	Value
Python Version	3.8+
Required Packages	requests, matplotlib
Optional Packages	pandas (for CSV manipulation), numpy (for statistics)

API Configuration

API	Endpoint	Rate Limit	Auth
OpenAlex	`https://api.openalex.org`	10 requests/second with polite pool (mailto header)	None required, but mailto header recommended
Semantic Scholar	`https://api.semanticscholar.org/graph/v1`	1 request/3 seconds without API key	Optional API key for higher limits

Important: Include a mailto header in OpenAlex requests to access the polite pool (10 req/sec). Without it, rate limits are lower.

3. Phase 1: 58-Corpus Construction

Location: literature/sections/introduction/

Step 1: OpenAlex Search (15_openalex_literature_search.py)

Execute 19 targeted queries against OpenAlex API

Search Queries (19 total)

Organized into 5 tiers by specificity:

"mutual fund style drift"
"fund style misclassification"
"style consistency" AND "mutual fund"
"investment style" AND "fund" AND "deviation"
"closet indexing"
"active share" AND "mutual fund"
"benchmark mismatch" AND "fund"
"style timing" AND "fund"
"window dressing" AND "mutual fund"
"return-based style analysis"
"holdings-based style analysis"
"Sharpe" AND "style analysis"
"style box" AND "Morningstar"
"fund manager incentives" AND "risk"
"tournament" AND "mutual fund"
"flow-performance" AND "mutual fund"
"fund misrepresentation"
"alpha" AND "style" AND "fund"
"fund classification"

Stage 1 Filters (Relevance)

Filter	Value
year_range	1990-2025
type	journal-article
language	English
has_doi	True
has_abstract	True

Stage 2 Filters (Quality)

Filter	Value
citations_min	10
or_top_journal	True
or_recent_with_3_cites	2020+ with 3+ citations

Outputs

raw_results.json (5400+ papers)
stage1_results.json (after relevance filters)
stage2_results.json (1594 papers after quality filters)

Step 2: Corpus Reduction (16_corpus_reduction.py)

Apply finance journal and keyword relevance filters

Stage 3: Journal Filter

Finance/Economics journals only

Exception: SSRN papers with 50+ citations pass

Stage 4: Keyword Relevance Filter

Papers must contain at least one of these terms in title or abstract:

style drift style consistency style timing closet index closet indexing active share benchmark mismatch misclassif* misrepresent* window dressing return-based style returns-based style holdings-based style style analysis style box

Context-dependent: investment style - "fund" must also appear in text

Step 3: Metadata Enrichment (17_enrich_corpus_metadata.py)

Fetch full metadata from OpenAlex and Semantic Scholar

Full abstract text
OpenAlex concepts/topics
Author affiliations
Reference counts
Open access status

Result: 58 papers with 100% relevance rate (all passed keyword filter by construction)

4. Phase 2: 53-Corpus Construction

Location: literature/scripts/

Methodology

Broad search filtered by publication in top finance journals:

Top 12 Journal Filter

Journal of Finance
Journal of Financial Economics
Review of Financial Studies
Journal of Financial and Quantitative Analysis
Financial Analysts Journal
Journal of Portfolio Management
Journal of Banking & Finance
Management Science
Journal of Monetary Economics
Review of Finance
Journal of Financial Markets
Journal of Corporate Finance

Scripts

Script	Output
`01_openalex_search.py`	raw_papers.csv
`02_supplementary_search.py`	merged_corpus.csv
`05_relevance_filter.py`	final_corpus.csv (53 papers)

Issue: No explicit keyword relevance filter - includes tangentially related papers

5. Phase 3: Merging and Classification

Location: literature/scripts/

Step 1: Merge Corpora

Source	File
58-Corpus	`literature/sections/introduction/openalex_output/enriched_corpus.json`
53-Corpus	`literature/data/final_corpus.csv`

Deduplication: DOI-based matching (case-insensitive)

Result: 111 unique papers (0 duplicates found)

Step 2: Keyword Classification (21_relevance_classifier.py)

Apply Stage 4 keywords to all merged papers

Classification Algorithm

def classify_paper(title: str, abstract: str) -> str:
    text = (title + " " + abstract).lower()

    # Check direct keywords
    for keyword in RELEVANCE_KEYWORDS:
        if keyword in text:
            return "RELEVANT"

    # Check context-dependent keywords
    if ("fund" in text or "mutual" in text):
        for keyword in CONTEXT_KEYWORDS:
            if keyword in text:
                return "RELEVANT"

    return "NEEDS_REVIEW"

Step 3: Generate Outputs

classified_papers.json (111 papers with classification)
final_relevant_corpus.json (71 papers)
final_relevant_corpus.bib
final_corpus_apa.html
corpus_statistics.json
corpus_statistics.html
charts/ (5 PNG visualizations)

6. Relevance Keywords Reference

Direct Keywords (17 terms)

Match triggers RELEVANT classification immediately:

style drift style consistency style timing active share closet index closet indexing return-based style returns-based style holdings-based style style analysis style box misclassif misrepresent window dressing fund style style rotation style risk

Context-Dependent Keywords (4 terms)

Only match if "fund" or "mutual" also appears:

investment style style deviation benchmark deviation style shift

Keyword Match Frequency (Final Corpus)

Keyword	Papers Matched
investment style [+fund]	20
style analysis	12
active share	11
fund style	6
window dressing	6
style drift	5
closet index/indexing	9
return-based style	4
misclassification	3
style timing	3

7. Final Results and Validation

65

Total Papers

6,580

Citations

101.4

Mean Cites

27

Journals

Corpus Composition

Source	Relevant	Total	Rate
58-Corpus (targeted search)	58	58	100%
53-Corpus (journal prestige)	7	53	13.2%

Validation Checks

All 71 papers have DOIs
All 71 papers match at least one relevance keyword
No duplicates exist in final corpus
Year range is within 1990-2025
All papers are from finance/economics journals or SSRN with 50+ cites

8. The Seven Additional Papers

These papers from the 53-corpus matched relevance keywords and were added to the final corpus:

#	Title	Journal	Cites	Matched Keyword
1	Mutual Fund Styles	Journal of Financial Economics	490	fund style
2	On Mutual Fund Investment Styles	Review of Financial Studies	422	investment style [+fund]
3	Liquidity, Investment Style, and the Relation between Fund Size and Fund Performance	Journal of Financial and Quantitative Analysis	332	investment style [+fund]
4	Active Share and Mutual Fund Performance	Financial Analysts Journal	296	active share
5	Mutual Fund Misclassification: Evidence Based on Style Analysis	Financial Analysts Journal	162	style analysis, misclassification
6	Equity Style Timing	Financial Analysts Journal	69	style timing
7	Diseconomies of Scale in Quantitative and Fundamental Investment Styles	Journal of Financial and Quantitative Analysis	13	investment style [+fund]

9. Complete Execution Guide

Execute these commands in order to fully reproduce the 65-paper corpus:

Phase 1: 58-Corpus Construction

cd literature/sections/introduction/

python 15_openalex_literature_search.py
python 16_corpus_reduction.py
python 17_enrich_corpus_metadata.py
python 18_generate_apa_html.py

Phase 2: 53-Corpus Construction (if starting fresh)

cd literature/scripts/

python 01_openalex_search.py
python 02_supplementary_search.py
python 05_relevance_filter.py

Phase 3: Merging and Classification

cd literature/scripts/

python 21_relevance_classifier.py
python 22_final_corpus_apa_html.py
python 23_corpus_statistics.py

Expected Final Outputs:

classification_output/final_relevant_corpus.json - 71 papers
classification_output/final_corpus_apa.html - APA citations
classification_output/corpus_statistics.html - Statistics dashboard
classification_output/charts/ - 5 PNG visualizations

Mutual Fund Style Drift Literature Corpus

Table of Contents