The effects of monetary policy in New Zealand : evidence from the SVAR analysis

Abstract

The primary objective of the research in this thesis is to examine empirically the effects of monetary policy in New Zealand. This is based on an econometric analysis of ways in which a monetary policy shock affects key variables in the New Zealand economy. To achieve this objective, this thesis formulates a parsimonious macroeconometric model using the well known Contemporaneous (Short-Run) SVAR techniques. Past empirical research on the effects of monetary policy in closed and open economies found evidence of several anomalies, such as the liquidity, price, exchange rate and forward discount bias puzzles. To resolve these puzzles, this thesis constructs a Contemporaneous SVAR model (called NZSVAR model) and formulates identification scheme(s), i.e. the NZ identification scheme, that lead to computation of impulse response functions that are free of these empirical anomalies. The results show that we appear to have successfully modelled monetary policy effects in New Zealand. In our analysis, we do not encounter those anomalous responses often documented in previous literature. Specifically, we have found evidence of neither the price nor the exchange rate puzzle. Furthermore, we do not observe significant increases in the level of output.

This thesis also conducts further investigation into two important issues. The first issue involves the starting values employed during the estimation process. An alternative approach to using zero starting, which is to use the estimates computed based on the recursive identification scheme, is introduced and recommended. This approach is considered particularly appropriate in our context because the NZ identification scheme was formulated based closely on the recursive scheme. The statistical results obtained appear to be robust across the sets of starting values so that we can be confident that we have achieved (at least one of the) global maxima.

The second issue involves the importance of including the oil price variable in the model. Based on the evidence from forecast error variance analysis, it was concluded that the world oil price variable might not be particularly useful in the New Zealand context. An innovation to the variable provides very little contribution to the variation of any of the domestic variables. Moreover, to investigate the issue further, we construct a 6-variable non-recursive SVAR model where the oil price is dropped but the model is identified in the same manner as the 7-variable NZSVAR model. As expected from the variance decomposition results, omission of the oil price variable does not have significant impacts on the point-estimate shapes and the associated error bands of the impulse response functions. However, to explain why we do not observe evidence of the price puzzle, this thesis argued that there are at least two other variables in the model that seem to be performing as indicators, which can capture forward-looking monetary policy actions. These variables are the New Zealand domestic short-term interest rate and the exchange rate.