Revealing ownership structure of funds using minimum spanning tree

Abstract

The traditional method of analysing a time series of stocks and funds is to use simple Pearson correlations. However, experience shows that cross-correlations are not an accurate indicator of the mutual ownership relations. We show that the minimum spanning tree methodology, previously used to perform more comprehensive studies of asset returns correlations, can be used to deduce the underlying ownership structure with reasonable accuracy. We also show that adjusting the time series for a common trend of stocks and subsequent filtering of the short term variations of returns using the ARIMA model is a prerequisite for this application of the minimum spanning tree.