Publication Date: October 2000
Recent time series methods are applied to the problem of forecasting New Zealand’s real GDP. Model selection is conducted within autoregressive (AR) and vector autoregressive (VAR) classes, allowing for evolution in the form of the models over time. The selections are performed using the Schwarz (1978) BIC and the Phillips-Ploberger (1996) PIC criteria. The forecasts generated by the data-determined AR models and an international VAR model are found to be competitive with forecasts from ﬁxed format models and forecasts produced by the NZIER. Two illustrations of the methodology in conditional forecasting settings are performed with the VAR models. The ﬁrst provides conditional predictions of New Zealand’s real GDP when there is a future recession in the United States. The second gives conditional predictions of New Zealand’s real GDP under a variety of proﬁles that allow for tightening in monetary conditions by the Reserve Bank.
Automated modeling, forecasting, PIC model selection, policy analysis, real GDP
JEL Classification Codes: C11, C22, C32, C53
See CFP: 1020