A regime change is a significant change in the collective trading behaviour in a financial market. Being able to detect the occurrence of regime change could lead to a better understanding and monitoring of financial markets. In this paper, a novel method is proposed to detect regime change, which makes use of a data-driven approach, that of directional change (DC). Compared to the conventional approach of using time series analysis, DC is an alternative approach to sample price movement. As variables observed under time series do not apply to DC, our first contributions is the identification of a new relevant indicator for regime change detection. Our second contribution is the comparison of both the DC approach and time series analysis, their ability to achieve regime change detection. The ability of both approaches in regime change detection is examined over a period of market uncertainty, that of Brexit. The results demonstrate that the DC approach is as effective as the time-series approach in detecting regime changes. Moreover, the DC approach is encouraging because some market regime changes are detected under DC, but are not found under time series. That means they support each other in the detection of regime change, and can also provide extra information to complement each other. Together, regime changes detected under both DC and time series provide a better insight into the market, which market participants and regulators could benefit from.
When financial market conditions change, traders adopt different strategies. The traders’ collective behaviour may cause significant changes in the statistical properties of price movements. When this happens, the market is said to have gone through “regime changes”. The purpose of this paper is to characterise what is a “normal market regime” as well as what is an “abnormal market regime”, under observations in Directional Changes (DC). Our study starts with historical data from 10 financial markets. For each market, we focus on a period of time in which significant events could have triggered regime changes. The observations of regime changes in these markets are then positioned in a designed two-dimensional indicator space based on DC. Our results suggest that the normal regimes from different markets share similar statistical characteristics. In other words, with our observations, it is possible to distinguish normal regimes from abnormal regimes. This is significant, because, for the first time, we can tell whether a market is in a normal regime by observing the DC indicators in the market. This opens the door for future work to be able to dynamically monitor the market for regime change.