7.3 Exogenous assumptions
Before either the dynamic model or the steady-state model can be calculated, a number of the values of certain exogenous variables must be specified. Key exogenous variables are:[20]
- Export and import price tracks in foreign price terms
- Government consumption, government investment and government employment
- The NAIRU, average hours paid per week and labour force participation rate
- Productivity variables
- Desired debt level of households
- Population growth, and
- Net foreign debt target.
Export and import price tracks in foreign price terms, as well as the government expenditure track, are supplied by members of the Treasury forecasting team responsible for the external and government sectors respectively. The exogenous nature of foreign prices is consistent with New Zealand's small open economy.
The time path for the NAIRU, the participation rate, productivity variables and the desired debt level of households are determined by filtering the data. Filtering is a process of using statistical methods either to decompose a historical time series into its trend and cyclical components, for a series where the variable of interest is observable; or, in the case where the variable is not observable (ie, not measured; for example, the NAIRU),[21] use other observable variables to estimate their value. For observable variables we can then use the recent trend of the variable of interest as the basis for inputting into the model as the future track of that variable in the forecast period.[22] Figure 14 shows the filtered trend over history for the labour productivity series.
- Figure 14: Estimation of the historical trend of the labour productivity series
Notes
- [20]A full list of exogenous variables is available in the Appendix.
- [21]The process for determining the NAIRU is documented in Guy and Szeto (2004).
- [22]We described the recent trend as being the basisfor the forecast. We used the term “basis” as sometimes the trend is adjusted in the forecast period to ensure that forecast trends and end points (ie, where the track ends up at the end of the forecast period) are not too dissimilar to previous forecasts. The rationale for this is we do not want large changes in the forecasts because of radically different estimation of trends; especially given the well-known sensitivity of filtering techniques to the end point of the sample they are estimated over.
