Improving the accuracy of bubble date estimators under time-varying volatility

Abstract

In this study, we consider a four-regime bubble model under the assumption of timevarying volatility and propose an algorithm of estimating the break dates with volatility correction. First, we estimate the emerging date of the explosive bubble, its collapsing date, and its recovering date to the normal market under the assumption of homoskedasticity. Second, we collect the residuals and then employ the weighted-least-squares-based estimation of the bubble dates. Using Monte Carlo simulations, we demonstrate that the accuracy of the break date estimators improves significantly via this two-step procedure in some cases compared to those based on the ordinary least squares method.

Full document