Document Type : Research Paper
Authors
1 Assistant professor of Accounting in department of management and Accounting Shahid beheshti university
2 MSc. Student of accounting, Management and Accounting Faculty, Shahid Beheshti University, Tehran, Iran
3 Ph.D. student, Department of accounting, Faculty of management and accounting, Shahid Beheshti University, Tehran, Iran
Abstract
Abstract
Purpose: Citation counts have long been seen as a marker of scientific credibility, but they often push researchers to pile up references for prestige rather than real relevance. This study looks at citation habits in Iranian accounting and auditing research through one clear question: how well do cited works match the main topic of the paper that cites them? To answer this, we create a new measure called thematic cohesion—the semantic similarity between a paper’s core idea and the titles of all its citations—and apply it systematically for the first time to Iranian accounting literature. The goal is to reveal the gap between citation prestige and actual usefulness. Iranian accounting scholars now publish internationally and feel strong pressure to cite high-impact journals. While this boosts visibility, it can weaken rigor if references are decorative rather than essential. We aim to quantify how relevant citations really are, build an automated tool to check citation quality, and move the focus from how many citations a paper has to how well they support its argument. By setting baseline cohesion levels and spotting differences across journals and years, we provide a foundation for better citation practices.
Methodology: We analyzed a full decade, 2014 to 2023, covering every article in Iran’s 14 machine-readable accounting journals approved by the Ministry of Science. This gave us 2,938 full-text papers. Using custom web-scraping tools, we pulled 132,543 references from their bibliographies. After strict filtering to keep only foreign journal articles—excluding Persian sources, conferences, books, and theses—we ended up with 65,146 valid citation pairs. Each cited journal was linked to its annual Scopus quartile to track prestige separately from relevance. Thematic cohesion was calculated as the cosine similarity between the title of the citing paper and the title of each cited work. We used a lightweight Sentence-BERT model trained for semantic similarity, producing scores from 0 to 100 percent. These were grouped into five bands: weak (below 35%), moderate (35–45%), good (45–55%), high (55–65%), and excellent (above 65%). Averages were computed for each journal and year, and the full pipeline—from scraping to scoring—is openly shared online so anyone can replicate or adapt it.
Findings: The results show a field that is strong on prestige but only moderate on relevance. Iranian authors heavily favor top-tier international journals: 77.5 percent of foreign citations go to Q1 outlets, 15.2 percent to Q2, 5.8 percent to Q3, and just 1.5 percent to Q4. The most-cited journals match global leaders: The Accounting Review, Journal of Accounting and Economics, and Journal of Accounting Research together account for over 14 percent of all citations. This confirms that Iranian scholars are fully engaged with the global prestige system. But when relevance is measured, the average thematic cohesion across all 65,146 pairs is 45.4 percent—solidly in the “good” range. The scores form a smooth bell curve centered between 45 and 50 percent, with a standard deviation of 12.3 percent. Only 9 percent of citations reach high or excellent levels above 55 percent, while 10 percent fall into weak territory below 35 percent. About 42 percent sit in the moderate band, meaning many references are loosely rather than tightly connected to the paper’s focus. At the journal level, Professional Auditing Research tops the list with 47.93 percent average cohesion, followed by Empirical Studies in Accounting at 46.98 percent and Accounting and Social Interests at 46.75 percent. Even the lowest performer, Accounting Advances, manages 42.82 percent—still moderate, but clearly behind the leaders. Over time, cohesion fluctuates little: it peaked in 2017 at 47.61 percent and dipped to 44.75 percent in 2021, always staying between moderate and good. This stability shows a settled citation culture, but the persistent ceiling below 50 percent suggests topical drift is routine.
Conclusion: These patterns matter because raw citation counts reward volume, not depth. Journals unintentionally encourage “citation inflation” by valuing quantity. An automated cohesion check offers a cheap, fair way to push back. We suggest a simple rule: require an average cohesion of at least 45 percent during peer review. This wouldn’t block creative or interdisciplinary references—it would just encourage authors to choose citations that truly strengthen their case. Editors who adopt this join a wider shift toward treating citation quality as its own standard. The picture that emerges is clear: Iranian accounting research cites the right journals, but not always the right ideas within them. The heavy lean toward Q1 sources proves strategic alignment with global norms. Yet a 45.4 percent average cohesion means most citations share less than half their meaning with the citing paper. That’s acceptable, but not ideal. Core references should cluster in the high or excellent range; moderate scores should be for background or methods, not the backbone of the argument. The bell-shaped distribution centered at 45–50 percent shows that “safe but loose” citing is standard practice. Authors pick respected works in the broad accounting domain even when the fit isn’t sharp. This makes sense under current rules—impact factors and citation totals drive careers—but it blunts precision. The narrow yearly range and small journal differences suggest cohesion is a shared habit, not shaped much by individual editorial policies. In short, the field has mastered prestige. The next step is relevance. By making thematic cohesion visible and measurable, this work gives editors, reviewers, and authors a practical tool to improve scientific integrity without adding heavy burdens.
Contribution: The study brings three fresh contributions. First, it defines and tests thematic cohesion—a direct measure of citation relevance, not just reputation. Second, it delivers a complete, automated workflow using Sentence-BERT that works at scale, even in non-English settings, with all code freely available. Third, it offers a ready-to-use peer-review filter: a quick cohesion report that flags papers needing tighter references. At almost no cost, journals can start rewarding substance alongside status. Together, these advances provide a blueprint any field or country can follow to move from counting citations to understanding them.
Keywords
- citation analysis
- natural language processing
- machine learning
- text mining
- Bidirectional Encoder Representations from Transformers (BERT)
Main Subjects
)