PathOS Impact of Open Access Routes on Topic Persistence Case Study Research Data, Code, and Analysis Results

This repository contains the data, scripts, and results for the Impact of Open Access Routes on Topic Persistence case study, part of the PATHOS project.

Overview

Artificial intelligence methods are being rapidly mobilized to tackle the climate crisis, but the knowledge base often burns bright and fades quickly. This case study asks whether two distinct Open Access (OA) routes help AI-for-Climate research topics stay active in the literature:

Green OA: self-archiving in repositories

Published OA: journal-mediated open access with a clear licence

Bronze OA and dual-mode publications are excluded for treatment clarity. Closed Access (CA) articles serve as the counterfactual.

By foregrounding topic persistence as a key dimension of impact, the study goes beyond short-term citation counts and investigates whether openness helps research topics remain visible long enough to demonstrate their potential.

Repository Structure

├── README.md
├── fos_taxonomy_v0.1.2.json
├── persistent_topics_create_collection.py
├── persistent_topics_find_paper_openaireids.py
├── persistent_topics_find_paper_affiliations.py
├── persistent_topics_get_collection_author_gender.py
├── persistent_topics_calculate_indicators.py
├── persistent_topics_calculate_indicators_sdg.py
├── persistent_topics_indicators_create_data_for_vis.py
└── persistent_topics_collection_w_outcomes/
├── complete_collection_df.parquet / .xlsx
├── topic_attribution_df.parquet / .xlsx
├── results/
│ ├── analysis_conclusions.txt
│ ├── summary_statistics.xlsx
│ ├── treatment_effects_green_oa.xlsx
│ ├── treatment_effects_published_oa.xlsx
│ ├── descriptive_effects_any_oa.xlsx
│ ├── tables/
│ │ ├── 01_executive_summary.xlsx
│ │ ├── 02_treatment_group_characteristics.xlsx
│ │ ├── 03_causal_effects_summary.xlsx
│ │ ├── 04_topic_persistence_analysis.xlsx
│ │ ├── 05_gender_equity_outcomes.xlsx
│ │ ├── 06_economic_impact_analysis.xlsx
│ │ ├── 07_publication_year_analysis.xlsx
│ │ └── 08_robustness_analysis.xlsx
│ ├── visualizations/
│ │ ├── 01_sample_overview.png
│ │ ├── 02_causal_effects.png
│ │ ├── 03_outcome_analysis.png
│ │ └── 04_temporal_and_balance.png
│ └── final_visualization_data_figures/
│ ├── data/
│ └── figures/
└── results_sdg_only/
├── sdg_analysis_conclusions.txt
├── green_matched_sdg_papers.xlsx
├── published_matched_sdg_papers.xlsx
├── closed_matched_a_sdg_papers.xlsx
├── closed_matched_b_sdg_papers.xlsx
├── tables/
│ ├── 01_sdg_distribution_matched_samples.xlsx
│ ├── 02_sdg_treatment_effects.xlsx
│ ├── 03_sdg_vs_non_sdg_comparison.xlsx
│ ├── 04_sdg_categories_by_impact.xlsx
│ ├── 05_sdg_gender_industry_collaboration.xlsx
│ ├── 06_sdg_analysis_summary.xlsx
│ ├── 07_sdg_alignment_comparison_matched.xlsx
│ └── 08_sdg_alignment_effects_summary.xlsx
└── visualizations/
├── 01_sdg_distribution_overview.png
├── 02_sdg_treatment_effects.png
├── 03_sdg_impact_analysis.png
└── 04_sdg_alignment_comparison_matched.png

Data Sources

External Data Sources (not included)

Semantic Scholar Academic Graph: full publication metadata

OpenAIRE Graph: European research infrastructure data

PATSTAT: patent database for citation analysis

ROR: Research Organization Registry

SciNoBo toolkit: FOS classification, interdisciplinarity, SDG mapping, FWCI scores

Included Data

Complete processed collection with outcomes

Topic attribution dataset (paper-topic mappings, persistence scores)

Analysis results: matched samples, treatment effects, summary statistics

SciNoBo Field of Science taxonomy (fos_taxonomy_v0.1.2.json)

Scripts

Data Processing

persistent_topics_create_collection.py – integrates multiple data sources, outcomes, affiliations, patent citations

persistent_topics_find_paper_openaireids.py – maps DOIs to OpenAIRE IDs

persistent_topics_find_paper_affiliations.py – extracts affiliations, science-industry collaboration

persistent_topics_get_collection_author_gender.py – gender classification of authors

Analysis

persistent_topics_calculate_indicators.py – main causal inference analysis (PSM for Green OA vs CA, Published OA vs CA)

persistent_topics_calculate_indicators_sdg.py – SDG-focused treatment effects

persistent_topics_indicators_create_data_for_vis.py – prepares final visualization datasets and figures

Key Findings

Sample

Total: 132,134 papers (2000–2021)

Green OA: 3,792 papers

Published OA: 19,045 papers

Closed Access: 92,998 papers

Contributions

New Topic Persistence Metric for long-term impact

Clean OA treatment definitions (excluding dual-mode and Bronze)

Separate analysis of Green vs Published OA pathways

Main Results

8 significant causal effects across outcomes

Enhanced topic persistence in OA papers

Positive gender equity outcomes

Evidence of economic impact (patents, collaborations)

SDG Findings

24,948 SDG-relevant papers (18.9% of sample)

11 significant treatment effects for SDG-related research

Stronger knowledge sustainability for achieving SDG goals

Methodology

Design

Propensity Score Matching (PSM) with balanced covariates

Separate analyses for Green OA vs CA and Published OA vs CA

Robust outcome metrics (including new persistence measure)

Treatment Definitions

Green OA: repository-based

Published OA: journal-based (gold, hybrid, diamond)

Closed Access: no open provision

Excluded: dual-mode and Bronze OA

Outcomes

Citation impact (traditional)

Topic persistence (novel metric)

Gender equity in authorship

Economic impact (patents, collaboration)

Field effects (disciplinary and SDG)