Bubbles for Fama from Sornette

Abstract

We present strong evidence that financial bubbles can be identified ex-ante and that a sharp price increase, when suitably qualified by a bubble indicator called LPPLS confidence, does on average predict unusually low returns going forward. For this, we use a methodology called log-periodic power law singularity (LPPLS) combined with the event study method applied to industry sectors in China over 2005–2020 and the US over 2009–2020. We identify a new class of apparent bubble regimes corresponding to the convergence to a stable price level, which can be disentangled using LPPLS-based indicators from standard bubbles followed by crashes.

Keywords: Bubbles, Log-periodic power law singularity, Confidence indicator, Bubble crash, drawdown

JEL Classification: C20, C40, C53, G01, G17