Complex Interactions in Financial Markets
All six studies show how complex the interactions on the micro-level (agent, bank, trader group) are and shed some light on how they determine the observed outcomes (price, network) on the macro-level. The two main findings in case of the artificial financial market are: (1) the introduction of a second asset triggers alternating herding behavior of the informed traders. This results from simple profit considerations by the individual agents rather than any form of coordination between them. (2) The FTT `bans' the informed traders already for very small tax rates and in contrast the FAT has almost no effect on them with respect to which market they submit their orders to. On account of this, the FTT is able to reduce the distortions from the fundamental value in the taxed market for very small tax rates, while the FAT is affecting the overall trading process less and therefore an introduction might bear less risk. Interesting future projects for an agent-based financial market could incorporate findings of behavioral finance, like overconfidence, anchoring and adjustment or home bias to observe how such fallacies work together in a market setup. For the banking network, the most important finding is that banks show persistent behavior over the complete sample (including the GFC) and strong preference for continued lending relations. Another interesting finding is that in terms of intermediate relations the high directed clustering coefficient (macro-level) seems not to be preferred by the banks, but emerges unintendedly out of their behavior which is steered by other preferences. However, these are only first steps and it is necessary to look for robust universal results for different banking networks and at the same time overcome the temptation of `nice' findings such as a power-law in the degree distribution which are observed in other disciplines. Moreover, a more profound knowledge of the behavior of banks is needed to finally tackle the task of modelling the financial system and provide reliable policy recommendations. In the investigation of the agricultural commodity markets it is interesting to see that the commodity index traders follow a clear `long only' strategy. They most likely seek to meet the demand of investors who want to include these assets for diversification purposes in their portfolio; a strategy that should make them relatively independent from the developments on these markets. Therefore, it makes sense that they neither systematically affect the price nor react to it. The `typical' trader groups of commercials and non-commercials on the other side also do not influence the price but react to it (in opposing directions). The results for energy commodities and precious metals show that the Money Managers (non-commercials) seem to Granger cause the most extreme price increase in the case of two out of six commodities, namely crude oil and silver. Hence, there is some evidence that excessive speculation might have influenced the prices of commodities. However, the regulators, in my view, should be primarily concerned with the less regulated off exchange trading than with futures markets. In order to add new insights to the existing literature in the future, a more detailed data set would be required.