Bubbles for Fama from Sornette

Galvanized by the claims of Greenwood et al. in Bubbles for Fama that “a sharp price increase of an industry portfolio does not, on average, predict unusually low returns going forward”, and Fama’s quote (June, 2016) that “Statistically, people have not come up with ways of identifying bubbles”, we present significant evidence to the contrary of both statements. Using a methodology called logperiodic power law singularity (LPPLS), which has been developed by the Sornette group over more than two decades, we show that a LPPLS-based “bubble confidence indicator” allows one to diagnose ex-ante the presence of a bubble. Using superposed epoch analysis, we find an excellent timing performance of price regime shifts, and more so, the larger the bubble confidence indicator. Moreover, we identify two classes of regime shifts following an accelerated price growth qualified by LPPLS: (i) bubbles followed by a large drawdown or crash, and (ii) price catch-up followed by a plateau, associated with the convergence to a stable price level. Indiscriminately mixing these two types of accelerated transient price increases may explain in part previous failures to diagnose bubbles and their aftermath. While the existence of the first class of transient accelerated price increases followed by crashes is a long-standing puzzle, the existence of the second class of transient accelerated price increases followed by a plateau poses a challenge to the efficient market hypothesis, thus constituting a new puzzle: the convergence to a stable price level, while accelerating, is slow, with investors and the market taking weeks to months to digest available information and to progressively converge to the final higher valuation consensus