On the Analysis of Irregular Stock Market Trading Behavior

Abstract

In this paper, we analyze the trading behavior of users in an experimental stock market with a special emphasis on irregularities within the set of regular trading operations. To this end the market is represented as a graph of traders that are connected by their transactions. Our analysis is executed from two perspectives: On a micro scale view fraudulent transactions between traders are introduced and described in terms of the patterns they typically produce in the market’s graph representation. On a macro scale, we use a spectral clustering method based on the eigensystem of complex Hermitian adjacency matrices to characterize the trading behavior of the traders and thus characterize the market. Thereby, we can show the gap between the formal definition of the market and the actual behavior within the market where deviations from the allowed trading behavior can be made visible. These questions are for instance relevant with respect to the forecast efficiency of experimental stock markets since manipulations tend to decrease the precision of the market’s results. To demonstrate this we show some results of the analysis of a political stock market that was set up for the 2006 state parliament elections in Baden-Wuerttemberg, Germany.