On trend robustness and end-point issues for New Zealand’s stylised business cycle facts

Abstract

We present new empirical evidence on trend robustness and end-point issues, utilising the macroeconomic data set investigated in McKelvie and Hall (2012). We consider the relative merits of non-robust Hodrick-Prescott (HP) and robust loess (LOcal regrESSion) trend filtering methods, and assess the sensitivity of HP1600 stylised facts to (i) the considerable “supply shock” deviations from trend associated with New Zealand’s 1992 power crisis, and (ii) an alternative HP100 specification and the loess approach. On end-point issues, we assess value-added from the use of seven-point triangular moving average and HP1600 filters, relative to insights from a 21-quarter uniform moving average filter.