Berlin 2018 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
SOE: Fachverband Physik sozio-ökonomischer Systeme
SOE 7: Poster
SOE 7.11: Poster
Montag, 12. März 2018, 17:00–20:00, Poster E
Clustering behaviour and long-range memory in a network-based financial market model with fitness-dependent preferential attachment — Florian Mix1,2, Julian Maluck1,2, and •Reik V. Donner1 — 1PIK Potsdam, Germany — 2HU Berlin, Germany
The Cont-Bouchaud model (CBM) is a classical network model of financial markets explaining heavy-tailed stock price fluctuations by the emergence of herding phenomena among traders. Here, we study a thorough extension of the CBM that replaces the uncorrelated random rewiring of links among traders by a simple dynamical process that accounts for the fitness of each trader (measured in terms of its respective economic performance in the past). Specifically, the network of traders exhibits a preferential attachment rule with the linking probability following a Fermi function in dependence on the difference between any two nodes' fitness values. We perform an extensive numerical analysis of the resulting model dynamics in terms of stock price, long-range memory of returns, and probability distributions of emerging cluster sizes, fitness and wealth. Unlike the fixed exponent of the cluster-size distribution of the classical CBM, our model allows for tuning the exponent in terms of the scale parameter of the Fermi function. However, we find that the (finite-time) power-law exponent and the maximum cluster size saturate (if at all) only for very long simulation times, indicating the presence of extraordinary long transients in the model.