Application of Bayesian and graph tools to a dynamic analysis of leadership relations among mutual funds

Financial literature has analyzed how previous trading decisions of some portfolio managers influence the subsequent trading of other peers. This paper contributes to the literature by proposing a new methodological approach to capture the time varying evolution of the leader-follower relationships. This approach allows capturing all the leader-follower relations that may occur during the sample period. To that aim, this work establishes a 4-steps procedure that combines Bayesian and Graphs tools. In the first step, the evolution of portfolios’ market exposures is estimated by means a Bayesian Capital Asset Pricing Model (CAPM). In the second step, a rolling Bayesian VAR model is used to dynamically estimate the paired leader-follower relationships. In the third step, graph tools are used to visualize the type and strength of these relationships, and the network topology of the industry. In the fourth step, we estimate and analyze the dynamic evolution of the leadership capacity of a portfolio. We illustrate the methodology with an application to Spanish mutual funds providing evidence that the relations between leader and follower funds are not constant along time. In addition, we study the factors that influence their leaderships finding that those funds with better performance and/or lower turnover are more likely to be leaders of other peers.