3  Identifying monetary policy

The specification for domestic interest rates (i) implied by Table 1 is intended to incorporate the monetary authority’s reaction function plus other shocks to domestic interest rates. Recent research suggests that for most of our sample period, the conduct of monetary policy in New Zealand approximates a Taylor-type (Taylor, 1993) reaction function (see Plantier and Scrimgeour, 2002; Huang, Margaritis and Mayes, 2001). The basic specification of the interest rate equation reflects this idea, but with several modifications.

The New Zealand monetary authority’s behaviour is probably appropriately described by forward-looking behaviour. This model therefore embodies a variant of the Taylor Rule in which the monetary authority reacts to forecasts of inflation and demand three quarters in the future. The Taylor Rule takes the form reflected in Table 1, but with the independent variables replaced with three-quarter-ahead forecasts generated from the reduced form VAR. Stock and Watson (2001) compare the implications of backward and forward-looking Taylor Rules in a three variable model of U.S. inflation. They found the choice of specification affected interest rate impulse response functions. Alternative specifications of the monetary authority’s reaction function applied in this New Zealand SVAR model revealed that the choice of a forward-looking specification produced different and more sensible reactions to interest rate shocks. Clarida, Gali and Gertler (1998) have also applied forward-looking Taylor Rules in an empirical context.

Over the sample period for this study there were several changes in the operation of monetary policy in New Zealand. Alternative specifications for the interest rate equation that included intercept and interactive dummy variables were tested to capture these changes. For example, a dummy variable that interacted with the exchange rate, and which took on a value of one between 1997:2 and 1999:4 and zero elsewhere, was used to capture the operation of monetary policy when monetary policy decisions were based on a monetary conditions index (MCI) (see Reserve Bank of New Zealand, 1996).[12] This alternative specification for the interest rate equation had little discernable effect on the impulse response functions used to identify monetary policy.[13]

Although the Official Cash Rate (OCR) is the current monetary policy instrument in New Zealand, this variable is unavailable for the full sample period. We have therefore used the 90-day rate as a proxy variable. The 90-day rate is not strictly controlled by the monetary authority and can be influenced by private expectations and shifts in portfolio decisions. We therefore include the world interest rate as a direct contemporaneous and lagged influence on the domestic interest rate.

The interest rate equation therefore takes the following form:

(7)    

where and are the three-quarter ahead reduced form model forecasts for the deviations from trend of log real domestic demand and log domestic prices.

Structural VAR models have typically been used to identify dynamic responses of an economy to particular shocks. This serves two purposes. It provides a means of analysing an estimated structural VAR and it reveals information about the dynamic properties of the economy investigated. The results can be used to inform policy makers and economic forecasters how economic variables such as real output and prices respond over time to changes in policy or other events.

As with all empirical work, the information value of dynamic simulations depends on the validity of the structure of the simulated empirical model. Because the focus of this paper is to measure the effect of monetary policy on New Zealand business cycles, of particular importance is that the model satisfactorily captures the dynamic impact of interest rates on the economy.

Impulse response functions have traditionally been used as a means of analysing an estimated structural VAR model (Hamilton, 1994). They represent the dynamic response of a variable in the model to an error term (referred to as a shock or innovation) in one of the structural equations. The transmission of the shock will depend on the form of the structural equations. Using Table 1, a shock to the domestic interest rate (i) will have a contemporaneous impact on the domestic exchange rate (e) and domestic asset returns (q) , and an impact on these and other variables one period into the future, two periods into the future,…, etc. These reactions represent the impulse responses.

Each variable in the model can be expressed as a combination of current and all past errors in the structural equations. That is, from equations (2) and (4), the SVAR can be written in moving average representation as follows:

(8)      y_t = A(L)^-1B₀^-1u_t = Θ(L)u_t

where Θ(L) contains the dynamic multipliers used to map out the impulse response functions following innovations to the structural error terms. The impulse response function represents the dynamic path for y_t from the ith equation following an innovation to the structural error term u_t from the jth equation, holding all other structural error terms constant.

A range of impulse response functions derived from this open economy SVAR model were used to evaluate the dynamic properties of the model and are illustrated in Buckle, Kim, Kirkham, McLellan, and Sharma (2002, Section 4, pages 19-27). The focus of this paper is the impact of monetary policy and therefore Figure 1 provides a selection of the impulse response functions from an interest rate shock.

Figure 1 shows the responses of several macroeconomic variables to an increase in the domestic interest rate. The size of the shock is an increase in the domestic interest rate of approximately 120 basis points above trend, which is equivalent to one standard deviation of the structural error term from the domestic interest rate equation. As is commonplace in the VAR literature, sixty-eight percent confidence bands have been estimated for the impulse response functions using the Monte Carlo bootstrapping approach of Runkle (1987). Both the impulse response functions and the sixty-eight percent confidence bands have been normalised by dividing by the size of the shock.

Monetary policy is expected to affect macroeconomic variables through a number of transmission channels. An increase in the domestic nominal interest rate will lead to an appreciation in the nominal exchanges rate, all else held constant. This appreciation in the nominal exchange rate results in a decrease in the domestic currency price of tradeable goods. When exporters are pricing goods in foreign currency terms, an increase in the nominal exchange rate results in a reduction in the domestic currency price of export goods. Likewise, an appreciation in the nominal exchange rate results in a decrease in the domestic currency price of imported goods. Changes in the domestic currency price of tradeable goods that are also consumed domestically, or that are used as an input in the production of goods that are consumed domestically, may take some time to be reflected in the domestic price of these goods. The magnitude and speed of exchange rate passthrough to the price of domestic consumption goods is dependent on a number of factors, including the level of domestic competition and fixed costs associated with altering domestic prices.

Figure 1 – Responses to domestic interest rate shock

Note: The solid lines represent the impulse response functions and the dotted lines represent the 68% confidence bands estimated using the Monte Carlo bootstrapping approach of Runkle (1987).

In the presence of short-run nominal rigidities, an increase in the nominal domestic interest rate results in an increase in the real domestic interest rate and an appreciation in the real exchange rate. A higher real exchange rate makes exporting and importing competing goods less competitive, resulting in lower aggregate demand. A higher real domestic interest rate reduces aggregate demand via a dampening effect on household consumption and firm investment.

Monetary policy may also affect aggregate demand and domestic prices via the credit channel. This channel arises from the impact of changes in the short-term interest rate on household balance sheets and bank lending decisions. For example, a higher domestic interest rate that results in a reduction in household equity may lead to lower housing investment, resulting in lower aggregate demand.[14] These effects are not identified in this model.

Figure 1 illustrates that an increase in the domestic interest rate (i) results in an immediate decline in domestic equity returns (q) as equities are substituted for bonds. Although equity returns (q) decline there appears to be little impact on domestic demand (d) (with the impulse response function not being significantly different from zero apart from the first quarter after the shock). The structural VAR model therefore suggests a weak transmission channel from domestic interest rates to domestic demand either indirectly through a reduction in equity returns or directly through a dampening effect on household consumption and firm investment.

The exchange rate (e) appreciates in response to the domestic interest rate (i) increase. The strongest reaction is after four quarters. The exchange rate (e) appreciation is likely to reflect the impact of an increase in the interest rate differential between domestic and foreign interest (i^w) rates. An appreciation of the exchange rate (e) results in a fall in export volumes (x), a decline in domestic output (y) and a fall in domestic consumer prices (p^c). While domestic prices start to fall almost immediately, the trough occurs over eighteen months after the shock. This is consistent with the presence of nominal price rigidities and slow exchange rate pass through (IMF, 2001). These impulses suggest the tradeables sector plays an important role in transmitting interest rate changes to domestic prices.

A number of VAR studies have encountered difficulties in detecting the impact of monetary policy actions on other macroeconomic variables. Kim and Roubini (2000) outlined several empirical “puzzles“ that have been associated with attempts to identify monetary policy in both open and closed economies. One attractive feature of this New Zealand SVAR model is that it does not encounter two of the puzzles discussed by Kim and Roubini (2000), namely the price puzzle (where the price level rises in response to a positive interest rate shock) and the exchange rate puzzle (where the exchange rate depreciates following a positive interest rate shock). Monetary policy appears to have been successfully identified, without the need to include non-monetary policy variables that several other VAR studies have had to resort to in order to identify monetary policy (see for example Brischetto and Voss, 1999 and Kim and Roubini, 2000).