The Importance of Being First: Position Dependent Citation Rates on arXiv:astro-ph

TL;DR

This study investigates how the position of e-prints in arXiv's astro-ph listings affects their citation counts, revealing a significant top-position advantage likely influenced by self-promotion and visibility.

Contribution

It provides the first quantitative analysis of position-dependent citation effects on arXiv astro-ph listings and tests potential explanations for this phenomenon.

Findings

Top-listed papers receive twice as many citations as lower-listed ones.

Self-promotion and visibility contribute to the positional citation effect.

Proposed mitigation strategies include randomizing listings and relevance-based sorting.

Abstract

We study the dependence of citation counts of e-prints published on the arXiv:astro-ph server on their position in the daily astro-ph listing. Using the SPIRES literature database we reconstruct the astro-ph listings from July 2002 to December 2005 and determine citation counts for e-prints from their ADS entry. We use Zipf plots to analyze the citation distributions for each astro-ph position. We find that e-prints appearing at or near the top of the astro-ph mailings receive significantly more citations than those further down the list. This difference is significant at the 7 sigma level and on average amounts to two times more citations for papers at the top than those further down the listing. We propose three possible non-exclusive explanations for this positional citation effect and try to test them. We conclude that self-promotion by authors plays a role in the observed effect but cannot exclude that increased visibility at the top of the daily listings contributes to higher citation counts as well. We can rule out that the positional dependence of citations is caused by the coincidence of the submission deadline with the working hours of a geographically constrained set of intrinsically higher cited authors. We discuss several ways of mitigating the observed effect, including splitting astro-ph into several subject classes, randomizing the order of e-prints, and a novel approach to sorting entries by relevance to individual readers.