Department of Mathematical and Computing Science, School of Computing, Tokyo Institute of Technology, Nagatsuta-cho, Midori-ku, Yokohama, Japan.
ETH Zürich, Department of Management, Technology and Economics, Zürich, Switzerland.
PLoS One. 2019 Aug 23;14(8):e0220645. doi: 10.1371/journal.pone.0220645. eCollection 2019.
Financial prices fluctuate as a results of the market impact of the flow of transactions between traders. Reciprocally, several studies of market microstructure have shown how decisions of individual traders or banks, implemented in their trading strategies, are affected by historical market information. However, little is known about the detailed processes of how such trading strategies at the micro level recursively affect future market information at the macro level. Using a special fined-grained dataset that allows us to track the complete trading behavior of specific banks in a U.S. dollar (USD) versus Japanese yen (JPY) market, we find that position management methods, defined as the number of units of USD bought or sold by banks against JPY, can be classified into two strategies: (1) banks increase their positions by trading in the same direction repeatedly, or (2) banks attempt to reduce their inventories by rapidly shifting their positions toward zero. We then demonstrate that their systematic position management strategies strongly influence future market prices, as demonstrated by our ability using this information to predict market prices about fifteen minutes in advance. Further, by detecting outlier trades, we reveal that traders seem to switch their strategies when they become aware of outlier trades. The evidence obtained here suggests that positions, which are a consequence of historical trading decisions based on the position management strategies of each bank, strongly influence future market prices, and we unravel how market prices at the macro level evolve through an interactive process involving the interaction between well-defined trading strategies at the micro level.
金融价格的波动是由于交易者之间交易流动对市场的影响。反过来，一些市场微观结构的研究表明，个体交易者或银行的决策如何受到历史市场信息的影响，这些决策被纳入其交易策略中。然而，对于微观层面的交易策略如何递归地影响宏观层面的未来市场信息，我们知之甚少。使用一个特殊的细粒度数据集，我们可以跟踪特定银行在美元（USD）兑日元（JPY）市场中的完整交易行为，我们发现，头寸管理方法（定义为银行买卖 JPY 的 USD 单位数量）可以分为两种策略：（1）银行通过反复进行同一方向的交易来增加头寸，或（2）银行试图通过快速将头寸转移到零来减少其库存。然后，我们证明他们的系统性头寸管理策略强烈影响未来的市场价格，正如我们使用这些信息提前预测市场价格十五分钟左右的能力所证明的那样。此外，通过检测异常交易，我们揭示了交易者似乎在意识到异常交易时会改变他们的策略。这里获得的证据表明，头寸是每个银行基于头寸管理策略的历史交易决策的结果，强烈影响未来的市场价格，我们揭示了宏观层面的市场价格如何通过微观层面上明确界定的交易策略之间的相互作用的互动过程演变。
