RDP 2008-01: A Sectoral Model of the Australian Economy 4. The Sectoral Effects of Monetary Policy

4.1 Expenditure Components of GDP

Figure 2 shows a panel of the estimated impulse responses of the domestic variables in our model to a monetary policy shock that generates an unexpected 25 basis point increase in the cash rate.^[11] According to our estimates, tighter monetary policy has a contractionary effect on domestic demand: residential investment, machinery & equipment investment, and consumption all fall below their baseline levels following an unexpected increase in interest rates. Imports also fall below their baseline level. Overall, we estimate that the 25 basis point tightening reduces real GDP below its baseline level by just over 0.2 percentage points, which is broadly in line with the estimates from other VAR studies (Figure 3).

The persistence of the GDP response to a monetary policy shock in our model is surprising given that monetary policy is generally thought to have a temporary effect on output. Although GDP does eventually return to its baseline level following a monetary policy shock, it takes longer to do so in our model than in most other Australian SVAR studies. This may reflect the fact that, at longer horizons, the precision of our estimation of the response of individual components to monetary policy shocks decreases. Evidence for this is provided by the line labelled ‘Six-variable SVAR’ in Figure 3. This shows the response of GDP to a monetary policy shock in an SVAR identical to ours in all regards, except that aggregate GDP, rather than its components, is estimated. In this smaller SVAR, GDP returns to its baseline level more rapidly than it does in the fully specified model of Equation (5).^[12]

Examining the sectoral effects more closely, dwelling investment and machinery & equipment investment respond most rapidly and strongly to the increase in interest rates; after four quarters both are around 0.6 per cent below their baseline level. Imports also respond to the shock quickly, though the effect is a little more moderate. Aggregate consumption falls away slowly and mildly after the increase in interest rates, while exports display little change. Of the three highly interest-sensitive components of activity, dwelling investment returns to its baseline level most quickly.^[13]

It is also worth noting the impact of the monetary policy shock on the other domestic variables in the model. The cash rate decreases quite rapidly after the initial shock, returning to baseline in about four quarters. Although the inflation rate exhibits a small price puzzle by rising initially, it falls below its baseline level two quarters after the shock.^[14] The real exchange rate does not initially appreciate in response to the monetary tightening (although the confidence interval is quite wide and spans zero). It appears to depreciate over the medium term, consistent with uncovered interest parity. As the responses of these non-GDP variables to a monetary policy shock are similar across the models that we study, we will not discuss them in detail in the remainder of the paper.

Although the sectoral results are broadly in line with our expectations, the fact that machinery & equipment investment is just as responsive to monetary policy as dwelling investment is interesting given that empirical studies in the US have found residential investment to be considerably more interest sensitive (Raddatz and Rigobon 2003; Bernanke and Gertler 1995). One possible explanation is that Australian firms may be more dependent on bank finance than their American counterparts (and so monetary policy may have a relatively strong effect on Australian business investment via the cash-flow channel). This certainly accords with the large contraction in machinery & equipment investment during the early 1990s recession in Australia when credit growth to businesses may have been constrained by supply-side considerations (Tallman and Bharucha 2000). The moderate effect of monetary policy on consumption makes sense when we remember that durable goods make up only 20 per cent of aggregate consumption. Note also that because consumption is a much larger share of aggregate activity than residential investment and machinery & equipment investment, it makes the largest contribution to the fall in GDP.

4.2 Production Components of GDP

We can also examine the effect of monetary policy across different industries. Because there are too many industries to include in a single SVAR, we estimate a separate model for each industry – an approach similar to that used by Bernanke and Gertler (1995). Equation (7) shows the identification scheme for each industry regression. This is similar to the expenditure components model, except there are only two domestic activity variables in each model – output in the industry of interest (Y_i,t), and output in all other industries (Y_t − Y_i,t). A potential problem with this approach is that the monetary surprises we estimate in each industry model may not be identical. However, our results suggest that structural errors in the interest rate equation are similar in each industry regression.

As with the expenditure components, we find evidence that output in some industries is more interest sensitive than in others (Figure 4). The largest response occurs in the construction sector, where output is 0.6 per cent below its baseline level 12 quarters after an unexpected 25 basis point increase in interest rates. Retail trade and manufacturing also decline, falling 0.4 per cent and 0.2 per cent, respectively, below their baseline levels 18 months after a contractionary monetary policy shock. These responses are broadly consistent with the results from the expenditure components model. In particular, the large decline in residential investment is consistent with the fall in production in the construction sector, while the relatively large response of retail trade probably reflects that sector's larger exposure to durable goods consumption than is the case for aggregate consumption. Among the least interest-sensitive industries are education, government administration, mining and agriculture, where supply-side factors are likely to be more important.

4.3 The Effect of a Shock to Domestic Consumption

These models of the Australian economy can also help us to understand the response of the economy to shocks to key variables besides interest rates. Using our expenditure SVAR, we first examine the impact of a shock to household consumption which, given its large share of aggregate GDP, is a potentially important source of shocks to the domestic economy. Figure 5 shows that, following an unexpected 1 percentage point increase in consumption, there is a noticeable expansion in residential investment and machinery & equipment investment while exports fall. This increase in domestic demand causes an increase in the rate of inflation and an appreciation of the real exchange rate.^[15]

Monetary policy reacts to the consumption shock, with the cash rate increasing quickly by 60 basis points. We can examine the effect this has on the other endogenous variables in the system by undertaking a simple counterfactual exercise: instead of allowing monetary policy to react endogenously to the initial shock to consumption, we hold interest rates constant at their baseline level. The dashed lines in Figure 6 show that when interest rates are held artificially low, all components of demand and inflation are higher than they would otherwise have been.

4.4 The Effect of a Shock to US Monetary Policy

Being a small open economy, changes in foreign economic conditions are likely to be a particularly important source of shocks to the Australian economy, and major downturns in the Australian economy have often coincided with major global downturns. In this section we examine the impact of a foreign shock on the expenditure components of Australian GDP.

To do this requires us to make some alterations to our baseline expenditure SVAR. With only two variables, the foreign block in our baseline model is too sparse to identify a meaningful foreign shock. Therefore, we expand our baseline model to include US inflation and the US federal funds rate, creating a four-equation foreign block.^[16] However, with these additional variables, and the six expenditure components of GDP, the model becomes too large to obtain reliable estimates. Hence, for this exercise we estimate a separate model for each expenditure component, as we did in our production components model.^[17] We chose to examine a US monetary policy shock, which appears to be reasonably well identified.

The impulse response lines in Figure 7 show the effect on the domestic economy of an unexpected 25 basis point increase in US interest rates. The increase in US interest rates has the expected impact on the foreign variables in the model – commodity prices, US output and US inflation all fall below their baseline levels. The decrease in foreign activity has a pronounced effect on the domestic economy – dwelling investment, machinery & equipment investment and consumption all decline. Inflation also falls, which leads interest rates to fall by 30 basis points four quarters after the shock. The real exchange rate depreciates after the shock, which cushions the impact of the shock on the traded-goods sector and contributes to an increase in exports, presumably to countries other than the US.

Replicating our simple counterfactual exercise, when domestic interest rates are held artificially high, all components of demand and inflation are lower than they would otherwise have been (results not shown). This suggests that over our sample period, Australian monetary policy has helped to dampen the impact of foreign shocks.

Footnotes

In response to a monetary policy shock, our 95 per cent confidence intervals span zero for most variables over most horizons. This is not an unusual result in the SVAR literature and reinforces the message that care should be taken in interpreting these results. As is standard in SVAR papers, our discussion focuses on the point estimates. [11]

Also, the sensitivity analysis of Section 4.1 suggests that monetary policy has a less persistent effect for a shorter, more recent sample period. [12]

As a check on the robustness of our results, we also estimate the impact of a monetary policy shock in models where we split domestic GDP into two components – a particular expenditure component of GDP and aggregate GDP less that expenditure component (as in Bernanke and Gertler 1995). The relative magnitude of the responses of the various expenditure components in these exercises is similar to our baseline model, although the response of machinery & equipment investment is now larger than the response of dwelling investment. [13]

The peak response of inflation is somewhat smaller than is found in some other studies (for example, Berkelmans 2005 and Nimark 2007), although it should be noted that the approach used in this paper is not intended to provide precise estimates of the response of aggregate relationships, but rather is intended to focus on differences across expenditure and/or production components of GDP. [14]

Although the shock to consumption should not be given an explicit structural interpretation (because other variables that influence consumption, such as employment growth and household wealth, are not included in the model), the fact that both output and inflation increase following the shock is consistent with a positive shock to demand. [15]

To identify a US monetary policy shock, we assume that the US monetary authorities have current-quarter information on commodity prices but only lagged information on output and inflation. [16]

Once again, the response of aggregate GDP, inflation, interest rates and the real TWI to the US monetary policy shock was similar across models. [17]