Ajouter au Calendrier
Université Paris 1 Panthéon-Sorbonne
(CV & Bio here)
A model of bubbles and crashes with non-local behavioral self-referencing
79 avenue de la République 75011 Paris
le vendredi 5 mai 2017
12:00 – 1:30pm, Amphi 4310
For security reason, please register before the deadline.
Deadline : 3 mai 2017
NB. If you are prevented from coming, we would be obliged if you could inform us as soon as possible at firstname.lastname@example.org.
Most existing models of financial bubbles and crashes, in particular the class of rational-expectation bubble models, derive the conditional expected return as being proportional to the contemporaneous crash hazard rate as a consequence of the standard risk-return relationship. We argue that the condition matching instantaneously return and risk is unrealistic and unlikely to be true in times of exuberant bubbles and of punishing crashes. We propose a class of models in which the hazard rate of jumps is determined by a non-local estimation of mispricing. Specifically, the mispricing is captured by a function of the difference between the present and the past prices over a long time scale, typically one year or more. This specification is rooted in behavioral finance, exploiting in particular the traits of « anchoring » on past price levels and on « probability judgement » about the likelihood of a correction as a function of the amplitude of the self-referential mispricing.
The insights obtained from numerical simulations of the model and estimation on market data are threefold : (i) In addition to the standard stylized facts, rising markets are understood as transient regimes when the risk of negative jumps is under-sampled while the investors expect a sufficiently large remuneration to compensate for the risk they anticipate. This makes quantitative the adage that « markets climb a wall of worry »; (ii) Reciprocally, the model cures a major problem of most crash jump models, which are in general rejected by data because they assume that crashes occur in a single large negative jump, by describing correctly that correction regimes and crashes are also phases with a significant duration, with inter-dependence between the sequences of corrections mediated by the interplay between the price and jump hazard rate dynamics; (iii) As a bonus, the model provides robust estimates of the risk premium, event in bearish markets, which is generally hidden. Our model provides a novel understanding of the risk-return relationship resulting from the entanglement of diffusion and jump risks.
|(Past and coming events)|