A network approach to unravel asset price comovement using minimal dependence structure

We develop a network representation-based methodology to aid an exploratory analysis of temporally evolving comovement in asset prices. This parsimonious order-n representation of the most significant comovement in asset prices, filtered by common factors, allows tackling a large number of assets and unraveling their complex comovement structure. Flexibility in choosing explanatory factors to suit the specific objectives of a study makes this methodology useful for portfolio analysis, risk parity approaches, and risk management decisions. We illustrate the features of the methodology for a set of major industry equity indices and to blue chip stocks, where we analyze the dynamic relevance of Fama-French factors. Investigating the network for more than 20 years, including the dot-com bust, global financial crisis, and European debt crisis, helps draw many insights. For instance, unexpected industries are seen to connect idiosyncratically through the dot-com bust. We demonstrate that a network factor model based portfolio allocation performs better than a regular factor model based allocation.