Detecting stock market turning points using wavelet leaders method

Abstract Detecting stock market turning points is a task with great significance and challenges. To achieve this purpose, we decompose the trend and cycle components of stock prices by the autoregressive fractionally integrated moving average model, which can simulate fractional difference stationary processes. What is more, we use wavelet leaders method to analyze multifractal characteristics of the cycle component and then propose two new indicators to detect the market turning points. Empirically, both indicators perform very well in detecting all critical turning points of US and China stock markets. Most importantly, compared with Bai et al. (2015) by testing the same data, our method detects turning points more accurately.