THE ECONOMETRICS OF MACROECONOMIC MODELLING

Responses to a permanent shift in interest rates

In this section, we discuss the dynamic properties of the full model. In the
simulations of the effects of an increase in the interest rate below we have not

Figure 9.9. Accumulated responses of some important variables to a 1 per
cent permanent increase in the interest rate RSt

incorporated the non-linear effect in the unemployment equation. Hence the results should be interpreted as showing the impact of monetary policy when the initial level of unemployment is so far away from the threshold value that the non-linear effect will not be triggered by the change in policy.

Figure 9.9 shows the simulated responses to a permanent rise in the interest rate RSt by 100 basis points, that is, by 0.01, as of 1994(1). This experiment is stylised in the sense that it is illuminating the dynamic properties of the model rather than representing a realistic monetary policy scenario. Notwithstanding this, we find that a permanent increase in the signal rate by 1 percentage point causes a maximal reduction in annual inflation of about 0.2% after three years.

Next, in Kolsrud and Nymoen (1998) it is shown that a main property of the competing claims model is that the system determining (w — p)t and (pi — p)t is dynamically stable. However, that prediction applied to the conditional sub­system, a priori we have no way of telling whether the same property holds for the full model, where we have taken account of the endogeneity of unemploy­ment, productivity, the nominal exchange rate, and the output gap (via the model of GDP output). However, the upper middle and lower left graphs show that the effects of the shock on the real wage growth, A(w — p)t, and on the change in the real exchange rate, A(pi — p)t, disappear completely in the course of the 24 quarters covered by the graph, which constitute direct evidence that stability holds also for the full system. The permanent rate of appreciation is closely linked to the development of the real exchange rate (v — p + pw)t: the increase in RSt initially appreciates the krone, both in nominal and real terms.

After a couple of periods, however, the reduction in Apt pushes the real exchange rate back up, towards equilibrium. Because of the PPP mechanism in the nominal exchange rate equation, the new equilibrium features nomi­nal appreciation of the krone, as Avt equilibrium corrects. This highlights the important role of nominal exchange rate determination—a different model, for example, one where Avt is not reacting to deviations from interest rate par­ity, would produce different responses. The two remaining graphs depict the response of the real economy. As real interest rates increase, aggregate demand falls and the unemployment rate ut increases, which dampens wages and prices.

9.2 Conclusions

The discussion in this chapter is aimed at several ends. First, as macroeconomic models typically are built up of submodels or modules for different parts of the economy, we have emulated this procedure in the construction of a small econometric model for Norway. Second, the chapter highlights the potential usefulness of such a model for the conduct of monetary policy. More specifically, we have argued that the success of inflation targeting on the basis of conditional forecasts rests on the econometric properties of the model being used.

Inflation targeting means that the policy instrument (‘the interest rate’) is set with the aim of controlling the conditional forecast of inflation 2-3 years ahead. In practice, this means that central bank economists will need to form a clear opinion about (and be able to explain) how the inflation forecasts are affected by different future interest rate paths, which in turn amounts to quant­itative knowledge of the transmission mechanism in the new regime. In this chapter, we show how econometrics can play a role in this process, as well as in an established regime of operational inflation targeting. In the form­ative period, the econometric approach will at least provide a safeguard against ‘wishful thinking’ among central bank economists, for example, that formally introducing an inflation target has ‘changed everything’ including the strength of the relationship between changes in interest rates and the overall price level. True, opting for inflation targeting is an important event in the economy, but one should take care not to overestimate its impact on the behavioural equations of a macroeconomic model that has given a realistic picture of the strength of the transmission mechanism over a sample that includes other, maybe equally substantive, changes in economic policy. Arguably, it may be a more robust procedure to regard at least the main part of the transmission mechanism as unaffected initially, and to take a practical view on the forecasting issue, that is using the model estimated on pre-inflation targeting data, and taking a practical approach to the forecasting issue, that is, using judgement and intercept corrections. Moreover, as experience with inflation targeting grows, and new data accumulate, the constancy of the model parameters becomes an obvious hypothesis to test, leading to even more information about how the economy operates under the new monetary policy regime.

We have presented a macroeconomic model for Norway, that we view both as a tool of monetary policy, and as providing a testing bed for the impact of the policy change on the economy. Conceptually, we partition the (big) simultane­ous distribution function of prices, wages, output, interest rates, the exchange rate, foreign prices, and unemployment, etc. into a (much smaller) simultaneous model of wage and price-setting, and several implied marginal models of the rest of the macroeconomy. The partitioning, and the implied emphasis on the modelling of a wage-and-price block, is anything but ‘theory-free’, but reflect our view that inflation in Norway is rooted in this part of the economy. More­over, previous studies—as laid out in this book—have established a certain level of consensus about how wage and price-setting can be modelled econo­metrically, and about how, for example, wages react to shocks to the rate of unemployment and how prices are influenced by the output gap. Thus, there is pre-existing knowledge that seems valuable to embed in the more complete model of the transmission mechanism required for inflation targeting.

In the previous study, Bardsen et al. (2003), based on data for the period 1966(4)-1996(4), valid conditioning of the wage-price model was established through the estimation and testing of the marginal models for the feedback variables, and—with one exception—they found support for super exogeneity of these variables with respect to the parameters in the core model. These results does not completely carry over to our current re-estimation of the core model on a dataset covering the period 1972(4)-2001(1). While the core model sustains broad specification tests, weak exogeneity no longer holds for the exchange rate with respect to the long-run parameters of the wage-price model. This implies a loss of estimation efficiency, which is only eliminated by simultaneous estimation of the core model together with the marginal models.

When we bring together the core model with the marginal models to the small econometric model for Norway, we show that the model can be used to forecast inflation. As regards the effects of monetary policy on inflation target­ing, simulations indicate that inflation can be affected by changing the short-run interest rate. A 1 percentage point permanent increase in the interest rate leads to 0.2 percentage point reduction in the annual rate of inflation. Bearing in mind that a main channel is through output growth and the level of unemployment, it is shown in Bardsen et al. (2003) that interest rates can be used to coun­teract shocks to GDP output. Inflation impulses elsewhere in the system, for example, in wage-setting (e. g. permanently increased wage claims), can prove to be difficult to curb by anything but huge increases in the interest rate.

Thus we conclude that econometric inflation targeting is feasible, and we suggest it should be regarded as a possible route for inflation targeters, alongside other approaches of modern open-economy macroeconomics.

This page intentionally left blank