The Treasury

Global Navigation

Personal tools

1  Introduction

Various economic models are used for forecasting the macro-economy at the New Zealand Treasury. One of the models, New Zealand Model (NZM) was a macroeconometric model of the NZ economy developed by Chris Murphy in 1995. In 1998, NZM was updated to allow for monetary policy operation under the Monetary Condition Index (MCI) regime. NZM was constructed in the style of the Murphy Model of Australia (see Powell and Murphy 1997), but modified to allow for differences in institutional structures, data availability and data properties between New Zealand and Australia.

The development of the macroeconometric model is an on-going process to incorporate latest policy changes, structural shifts in the data, and the development of economic theory and modelling. Over recent years, Treasury has re-developed the core macroeconometric NZM model. This has resulted in a new model reported in this paper. We refer to this new model as New Zealand Treasury Model (NZTM), to distinguish it from NZM.

The structure of the New Zealand Treasury Model (NZTM) differs from that the New Zealand Model (NZM) in three major areas. They are the relative price structure, the determination of the equilibrium exchange rate and the demand-pull framework that determines inflation. As a result of these developments, greater emphasis has been placed on using NZTM in aiding the Treasury forecast for the New Zealand economy.

One of the drawbacks of NZM was that the monetary policy reaction function was a contemporaneous price-level-targeting rule. Since the core theoretical structure was based on price levels, it was not possible to formulate the monetary rule as an inflation-targeting rule. Under a price-level target, following a shock to prices, the price level needs to return to the target level at some future point. Therefore, a price-level targeting rule tends to place a more demanding requirement on monetary policy than an inflation-targeting rule. In particular, price-level targeting means that the central bank needs to tighten or loosen harder and longer to achieve the target a price shock. This can be seen when comparing the dynamic response between two targeting rules. The variability of inflation and the variability of output are greater under price-level targeting than under inflation targeting because base level drift is not accepted under price-level targeting[1].

Since the core structure of NZM was based on price levels, the steady state model within NZM was simulated before running the dynamic version of the model in order to provide equilibrium values for forward-looking variables. Under a price-level targeting rule, all the price levels were well anchored in the long run. In this framework, the solution of the steady-state model on nominal exchange rates provided appropriate terminal conditions for the dynamic model.

In contrast, NZTM has a relative price structure that allows the monetary policy rule to be specified as an inflation target. Unlike NZM, NZTM does not have a fixed end point on the nominal exchange rate. Instead, the steady-state version of the model generates an equilibrium path for the real exchange rate that provides a key anchor for the dynamic structure of the model.

In NZM, production of the domestic good was demand-determined in the short run, with firms gradually adjusting the price of the domestic good to a medium-term target, which can be interpreted as the marginal cost of producing domestic goods. In this framework, the pass-through from wages and import prices into domestic inflation was strong. In NZTM, the pass-through from the labour market into domestic inflation has been reduced which might reflect both labour and product market reforms.

The production block of NZTM has been econometrically estimated (Szeto 2001). Hence, the steady-state version of the model provides an estimate of long-run potential output, which in turn provides the basis for a measure of the output gap used in the dynamic model. The output gap plays a major role in determining inflation in NZTM.

There is no single model that meets the requirements of both policy evaluation and forecasting. For policy simulation, more emphasis is placed on the theoretical structure of the model. For forecasting, more emphasis is placed on how well the specification of the model represents the data. It is also essential that a forecasting model is able to clarify how judgements about the key model assumptions affect the forecast. Furthermore, there is always a trade-off between complexity and simplicity. The main role of NZTM is to help the Treasury forecast the economy. With that role in mind, the theoretical structure of NZTM is parsimonious.

For example, estimates of the non-accelerating inflation rate of unemployment (NAIRU) are exogenously determined in NZTM. In theory, there are many factors determining the level of the NAIRU and those factors are not immutable. Therefore, for policy simulation, it is important to have a formal model of how the level of the NAIRU is determined. However, for forecasting, it is reasonable to assume that the level of the NAIRU is relatively constant over the short to medium-term forecast horizon.

In addition to providing the key anchor for the dynamic component of the model, the steady-state structure also strengthens the properties and analytic capabilities of the dynamic component. After a considerable number of time periods has elapsed following a shock, the dynamic adjustment path converges to a growth path consistent with the steady state solution. Accordingly, the steady state structure is used to study the equilibrium effects of permanent shocks and provides a quality assurance on the long run properties of the dynamic model.

The main purpose of this paper is to describe the structure of both the steady state and dynamic versions of NZTM. Section 2 describes the formal theory that supports the steady-state model. Section 3 describes the calibration of the model and the numerical steady state it attains. In Section 4 we describe a number of permanent shocks to exogenous variables that help to illustrate the steady state properties. Section 5 outlines the dynamic structure of NZTM while Section 6 presents a series of shocks from the dynamic model. Concluding comments and discussion of potential applications of the model are contained in Section 7.


  • [1]By contrast, Svensson (1999) found that price-level targeting results in lower short-run inflation variability than inflation targeting if output is at least moderately persistent.
Page top