A0850
Title: Influencer detection in market sectors via sparse network analysis
Authors: Simon Trimborn - University of Amsterdam (Netherlands) [presenting]
Kexin Zhang - City University of Hong Kong (Hong Kong)
Abstract: When financial market participants expect that news relating to a company is representative of other companies within the same sector, then the performance of that company on the markets is expected to drive the other assets' performance as well. Such situations often arise during earning announcement season. A Sparse Network Model (SNM) is introduced to identify the influential assets within sectors. Usually, sectors comprise a large number of assets relating to the issue of high dimensionality. Naturally, not all assets within a sector are expected to impact the performance of others; hence often, a sparse underlying structure arises. Sparse estimation techniques are often applied to uncover such structures. When particular structures like groups or blocks are part of the network, then a tailored estimator commonly provides a more accurate estimation. As such, an estimator is introduced to detect influencers in asset networks. The methodology is flexible. As such, it extends various sparsity estimation techniques towards detecting influencers when they are present. The asymptotic properties of the estimator are studied, and its performance in extensive synthetic data experiments is validated. The impact of assets is studied on others within the sectors of the S&P100. The aim is to illustrate which companies are most influential for the sector and document the dynamics in the influencer structure over time.
