RDP 2005-06: Credit and Monetary Policy: An Australian SVAR 4. Conclusion

This paper uses a structural vector autoregression (SVAR) model to examine the endogenous relationships between credit and other key macroeconomic variables, in particular monetary policy. Variance decompositions indicate that, at short horizons, shocks to the interest rate, the exchange rate, and past shocks to credit are found to be important in explaining movements in credit. Over longer horizons, shocks to output, inflation and commodity prices are found to play a greater role. For the domestic variables in general, the exogenous international variables are responsible for a large proportion of forecast errors at longer horizons.

The model suggests that a shock to the interest rate, increasing it by 25 basis points, results in the level of credit being almost half of a percentage point lower after four quarters. If monetary policy subsequently reacts in a manner consistent with its past behaviour, credit continues to decline for about four years, when it is almost 1 per cent lower than the counterfactual level. It then slowly retraces this decline. The timing of the response of credit appears to be similar to that of inflation; the response of output is more rapid, reaching the maximum response after about five quarters. The response of the other domestic variables accords with responses found elsewhere.

The SVAR estimates suggest that in response to a shock to credit, monetary policy plays an effective role in stabilising the economy. The impact of the credit shock on output and the exchange rate are almost completely offset by the response of monetary policy. Monetary policy does not completely counteract higher inflation, which is above baseline for about three years after the shock to credit. But inflation would be higher still over this period if monetary policy was passive to the macroeconomic consequences of the credit shock. Changes in credit are also moderated as a result of monetary policy's response.

The model is robust to changes in the lag length, but slightly less so to changes in the sample period. The sample period spans financial deregulation and the adoption of inflation targeting. It is then perhaps not surprising that the SVAR, which summarises the average dynamic properties of the data over the sample period, is somewhat sensitive to this. A further caveat is that the Killian-bootstrapped confidence intervals are relatively wide, making conclusive statements difficult. Nonetheless, the model presents plausible economic interactions, both in their timing and magnitude.