|Day & Time||27th October 2016|
|Venue||Hokkaido University Graduate School of Economics and Business Administration 3rd floor|
|Lecturer||Prof. Hiroshi Yamada (Hiroshima University)|
|Presentation Title||Quantile Hodrick-Prescott filtering|
|Abstract||Quantile regression was introduced in the seminal work by Koenker and
Bassett (1978) and widely applied in econometrics. Hodrick-Prescott (HP) (1997)
filtering is used frequently to estimate trend components of macroeconomic time series. In this paper, we contribute to the literature on macroeconometrics by introducing a filtering method that combines these two statistical tools.
We refer to it as quantile HP (qHP) filtering. qHP filtering enables us to obtain not only the median trend, which is more robust to outliers than HP filtering, but also other
quantile trends, which may provide a deeper understanding of time series properties.
As in the case of HP filtering, it requires selection of tuning parameter. We propose a method for selecting it, which enables us to compare trends from (q)HP filtering.
As an empirical example, we present some estimated quantile trends from Japan’s in dustrial index of production (IIP).