||This paper applies Log-Periodic Power Law (LPPL) model to Taiwan stock market to predict the regime-switching time of the 2008 bubble and crash. Moreover, this paper is dedicated to explaining the log-periodic price pattern with investor herding behaviors and granting the model a more intuitive financial interpretation. |
In contrast to the original methodology proposed by Johansen, Ledoit, and Sornette (2000), rather than the estimated range of critical time, we focused on the log-periodic price pattern (specifically, the log-periodic oscillation parameter), a crucial phenomenon before the crash as a prophetic sign for the crisis. Furthermore, to decipher the impact of herding on crash, we use a conditional probabilistic herding measure for the concept of deviation from market consensus, S-statistic, to distinguish anti-herding from herding by investor types. Finally, we construct a two-regime Threshold VAR model to examine the dynamic relations among herding, anti-herding and log-periodic price pattern.
To our surprise, the study finds that anti-herding behaviors of institutional investors strengthen the log-periodic oscillations while the effect is opposite for individual investors anti-herding behaviors. Institutional herding weakens the log-periodic pattern. This result may be counterintuitive, however, this result indicates that herding and anti-herding can truly reflect the complex mechanism of financial market before the crash and thus these behavioral factors are qualified as predictive indicators for financial crisis.