Title: Testing for serial correlation of unknown form using signed path dependence
Authors: Fabio Dias - University College London (United Kingdom) [presenting]
Franz Kiraly - University College London (United Kingdom)
Gareth Peters - Heriot-Watt University (United Kingdom)
Abstract: Whilst several tests for serial correlation in financial markets have been proposed and applied successfully in the literature, such tests provide rather limited information to construct predictive econometric models. This gap is filled by providing the following contributions: (i) a formal definition of signed path dependence based on how the sign of cumulative innovations for a given lookback window correlates with the future cumulative innovations at a given forecast horizon; (ii) theoretical results validating the definition on well-known model classes; (iii) a formal inference procedure to detect serial correlation of unknown form based on a hypothesis testing formulation of signed path dependence; (iv) experiments on synthetic data validating the test formulation via type I and type II error curves as functions of the significance threshold; (v) an application of the test on observed returns of global equity indices and currencies; (vi) a predictive econometric model for future market returns based on the concept of signed path dependence; and (vii) an application of this model as a profit-seeking trading strategy. It is found strong evidence of serial correlation of unknown form on equity markets, being statistically significant and economically significant even when trading costs are present.