Investigating the 'old boy network' using latent space models

TL;DR

This study uses latent space models to analyze the network of rugby fixtures among English schools, revealing how geographical proximity and institutional similarities contribute to the 'old boy network' phenomenon.

Contribution

It applies Bayesian latent position cluster modeling to uncover nuanced factors influencing school network formation, highlighting the role of fees and boarding proportions.

Findings

Geographical proximity strongly influences fixture networks.

Homophily in fees and boarding correlates with community formation.

Evidence of an 'old boy network' based on institutional characteristics.

Abstract

This paper investigates the nature of institutional ties between a group of English schools, including a large proportion of private schools that might be thought of as contributing to the 'old boy network'. The analysis is based on a network of bilaterally-determined school rugby union fixtures. The primary importance of geographical proximity in the determination of these fixtures supplies a spatial 'ground truth' against which the performance of models is assessed. A Bayesian fitting of the latent position cluster model is found to provide the best fit of the models examined. This is used to demonstrate a variety of methods that together provide a consistent and nuanced interpretation of the factors influencing community and edge formation in the network. The influence of homophily in fees and the proportion of boarders is identified as notable, with evidence that this is driven by a community of schools having the highest proportion of boarders and charging the highest fees, suggestive of the existence and nature of an 'old boy network' at an institutional level.