Ultrametricity in Hedge fund Selection and Fund of Funds Diversification

Gabriele Susinno and Augusta Miceli

Abstract:
Minimum market transparency requirements impose hedge fund (HF) managers to use the statement declared strategy in practice. However each declared strategy may actually origin a multiplicity of implemented management decisions. Is then the "actual "strategy the same as the "announced" strategy? Can the actual strategy be monitored or compared to the actual strategy of HF belonging to the same "announced" class? Can the announced or actual strategy be used as a quantitative argument in the fund of funds policy? We present how methods akin to the study of complex systems in Physics, Engineering, and Biology can be borrowed to enhance the investor's ability to see what is behind the historical evolution of a net asset value (NAV). With the appropriate metric, it is possible to draw a minimum spanning tree (MST) to emphasize the similarity structure that could be hidden in raw correlation matrix. We observe that some strategies effectively determine clusters of homogeneous announced strategy funds, while others don't. We believe this can be related to the degree of subjective discretionary decisions versus objective model decisions implied by the type of strategy: the formers create differences, while the others similarities.
Homogeneous clusters can be used as a reference for identifying anomalous funds. Clusters may be used to define coherent benchmarks. In a portfolio diversification process the most heavily connected node in a cluster can be used as the most representative of his own class. In general, as a highly graphical tool, fund trees in association with a solid due diligence, may facilitate the incorporation of subjective judgment in the investment selection process.