This paper has updated and extended previous studies that estimated the NAIRU for New Zealand using the Kalman filter on a reduced form equation. Extensions to previous studies include using the Stock and Watson (1998) procedure to estimate the signal-to-noise ratio; including a survey measure of inflation expectations as a variable; estimating the NAIRU for three different measures of inflation and two different models; including measures of short-term shocks based on their significance for New Zealand; and making use of quarterly data up to the end of 2003.
This paper found evidence to support the concept of a NAIRU with the unemployment gap found to be a significant variable in helping to explain the deviation between inflation and inflation expectations in the models estimated. The paper also found that the Kalman filter reduced form estimates showed a larger cyclical dimension to the labour market in the early 1990s than the HP filter, with the HP filter tending to follow the actual unemployment rate more closely.
The results indicate that the NAIRU is likely to be around the current level of the unemployment rate, which supports our prior expectations based on signs of a tight labour market. However, this paper has presented a range of estimates from different possible models, which all also have error bands around their estimates. These models can only provide indications of the current level of the NAIRU rather than point estimates, but also provide a useful guide to how the NAIRU has changed over time. All of the models indicate a NAIRU that is rising over the late 1980s and early 1990s, and then gradually falling with a brief respite around 2000 with the impact of Asian Crisis and two droughts.
This paper provides estimates of the NAIRU using a basic Kalman filter model. A number of extensions and areas of further work have been suggested that could not be included in this paper but may warrant further work. One area of work is to include the soil moisture deficit (Buckle et.al., 2002) as one of the supply shock variables because the omission of such a variable could result in a different estimate of the NAIRU’s variation.
A second area of work suggested is using nonlinear modelling to investigate whether the unemployment rate adjusts in a non-linear or asymmetric fashion, in which case the linear models used in this paper would be inappropriate (Skalin and Teräsvirta 2002). A third area is to investigate the performance of the various models by looking at the real-time versus ex-post properties of the various NAIRU measures (Graff 2004). Finally, adding a structural NAIRU equation to the models is an area which would be very challenging but if successful could help to understand the determinants of the NAIRU.