Xin Zhang (Sveriges Riksbank)

Score-driven tail shape, with application to bond yields at a high frequency (with Andre Lucas and Bernd Schwaab)

Abstract:
We propose an observation-driven modeling framework for estimating time series variation in the tail shape parameter of a Generalized
Pareto Distribution. We discuss different ways of handling non-tail observations and relate tail shape variation to observed covariates. We then
use the model to study the yield, volatility, and tail shape impact of bond purchases by the European Central Bank between 2010--2012. Bond
purchases lowered the conditional mean of bond yields by about -0.2 to -3.5 bps per 1 bn of purchases. The announcement of the program had a significant impact on the tail shape.