A0306
Title: Change point estimators with the weighted objective function when estimating breaks one at a time
Authors: Toshikazu Tayanagi - Hitotsubashi University (Japan)
Eiji Kurozumi - Hitotsubashi University (Japan) [presenting]
Abstract: The change point estimators are investigated with the weighted objective function proposed by a prior study when the model has two structural breaks and these breaks are estimated one at a time. The finite sample distribution of the first-step and second-step estimators are investigated under the long-span asymptotic scheme, which is valid when the break sizes are not so small. It will be shown that they are unimodal and asymmetric in general. However, for the small size of the breaks, the limiting distribution based on the long-span scheme cannot approximate the finite sample distributions well; the finite sample distribution of the first step LS estimator has four peaks while that of the first step weighted estimator tends to have two peaks at around the true break dates. It is shown that the latter property can be replicated based on the in-fill asymptotic scheme.
