5 Time-Varying Estimation

For all the models examined, plausible results have been obtained in terms of the direction of the domestic impacts of shocks to China and US growth. That is, the foreign output shock increases domestic New Zealand output and inflation, with monetary policy responding by increasing interest rates. The results of section 4 also imply that commodity prices are important for the transmission of these effects. Nevertheless, these results are obtained using a relatively long sample of data (25 to 30 years), whereas the correlation analysis in subsection 3.1 points to changing impacts of output growth on New Zealand over time.

To examine the possibility of time variation in the impacts of foreign shocks on New Zealand, we estimate the SVAR model with aggregate commodity prices, as used in subsection 4.2, but now estimating the coefficients over a moving sample of ten years of data. The focus of interest is China, and Figure 11 plots the estimated four-quarter (accumulated) impacts over time of a shock to China GDP on the other variables of the model. Note that, in common with the correlations in Figure 1, the values are plotted against the first quarter of the ten years over which the model is estimated.

These results strongly suggest that spillovers from China's growth to New Zealand GDP vary over time. In terms of GDP volumes, the estimated impact rises to around 1.1 for the decade from the mid-1990s. Although dropping to be close to 0.2 from the late 1990s, this rises again to about 0.3 at the end of the sample. This last result is in keeping with the discussion of Bowman and Conway (2013a, 2013b) about the importance of the Chinese market for New Zealand's agricultural exports in recent years. It is also striking in Figure 11 that China has an increasing impact on commodity prices from the early 1990s, with these factors also implying an increased impact on New Zealand's domestic interest rates and real exchange rate.

It is important to note that the focus of analysis is on the volume of GDP that does not reflect the income gains or losses from the terms of trade changes. To reflect these changes, we repeat the exercise by replacing GDP volumes with Gross National Expenditure (GNE). GNE measures New Zealanders' expenditure and therefore better reflects income changes owing to terms of trade movements. The results (red dotted line in Figure 11) show that the impact on GNE exhibits an upward trend during 1997-1999 reflecting the significant terms of trade gains in recent years.

Rather than using moving data windows, SVAR modelling sometimes takes account of time-variation in responses through applying a time-varying parameters VAR approach, as for example in Stock and Watson (2005). Although this approach is quite promising, the computational burden of estimating an eight-variable SVAR of this type would be substantial (Del Negro, 2003). Therefore, we do not pursue this here and leave it as an avenue for further research.

Figure 11: Time-varying four-quarter cumulative impact of China GDP shock

Notes: Values shown are the four-quarter accumulated impacts computed from the SVAR of subsection 4.2, estimated over a ten year moving window starting at the indicated date. Source: Authors' calculations.