Transmission channels and. model properties
In this chapter, we develop an econometric model for forecasting of inflation in Norway, an economy that recently opted for inflation targeting. We illustrate the estimation methodology advocated earlier, by estimating and evaluating a model of prices, wages, output, unemployment, the exchange rate, and interest rates on government bonds and bank loans. The model is built up sequentially. We partition the simultaneous distribution function into a small model of wages and prices, and several marginal models for the rest of the economy. The choice of model framework for the wage and price model follows from the analysis in earlier chapters. We use the model to analyse the transmission mechanism and to address monetary policy issues related to inflation targeting.
On 29 March 2001 Norway adopted inflation targeting. Rather than stabilising the exchange rate by pegging the Norwegian Krone to the Euro (or previously a basket of foreign currencies) the central bank became committed to an inflation target of 2.5%. This was in line with an international trend, as countries like Canada, New Zealand, Sweden, and the United Kingdom had already changed their monetary policy towards an explicit inflation target; cf. Bernanke et al. (1999).
Research on monetary policy has focused on the conditional inflation forecast as the operational target for monetary policy, yet the literature is dominated by either theoretical or calibrated models—examples are Ball (1999), Batini and Haldane (1999), Rpisland and Torvik (2004), Walsh (1999), Svensson (2000), Woodford (2000, 2003) and Holden (2003). True to the
approach taken in this book we will argue that econometric evaluation of models is useful, not only as an aid in the preparation of inflation forecasts, but also as a way of testing, quantifying, and elucidating the importance of transmission mechanisms in the inflationary process. In this way, inflation targeting moves the quality of econometric methodology and practice into the limelight of the economic policy debate.
Inflation is a many-faceted phenomenon in open economies, and models that include only a few dimensions, for example, the output gap and expectations of the future rate of inflation, are likely to fail in characterising the data adequately, as demonstrated in Chapter 7. Econometric work that views inflation as resulting from disequilibria in many markets fares much better (see Hendry 20016 and Juselius 1992). Our starting point is therefore that, at a minimum, foreign and domestic aspects of inflation have to be modelled jointly, and that the inflationary impetus from the labour market—the battle of markups between unions and monopolistic firms—needs to be represented, for example, as in the Incomplete Competition Model (ICM) which also stands out as the preferred model in Chapter 8.
The approach taken in this chapter to construct a small model of inflation is illustrated in Figure 9.1.
The focus is on the simultaneous wage-price model Dy(yt | zt, Yt-i, Zt-i), where yt = [wtpt]', the vector zt contains all conditioning variables, and (Yt-i, Zt-i) collects all lagged values of yt and zt. The variables in zt are partitioned into [z1t z2,t z3 t]', where z1t denote feedback variables, z2,t are non-modelled variables, and z3,t are monetary policy instruments. Lagged values are partitioned correspondingly, Zt-i = (ZM-i, Z2,t-i, Z3,t-i).
Figure 9.1. Model-based inflation forecasts
The feedback variables z1jt include unemployment, output, productivity, and import prices. Figure 9.1 indicates that the marginal models, Dzi (z1t | z2,t, Z3,t, Y-i, Zt-i), are not only functions of lagged wages and prices, but may also depend on both the non-modelled explanatory variables z2,t and on the policy variables z3 t. The feedback variables are treated as weakly exogenous variables in the wage-price model. This is a testable property that we address in Section 9.4 after modelling the feedback relationships.
The conditional non-modelled variables z2,t consist of domestic tax-rates and world prices. The crucial question for the policy instruments z3 t is whether there exists a single reaction function for the interest rate. Norway was pegging its exchange rate to different currency baskets throughout the sample period, which is 1972(4)-2001(1). For a substantial part of this time period the country saw frequent devaluations, particularly in the 1980s. Finding an empirically constant reaction function from inflation forecasts to interest rates is therefore a non-starter. Hence, we treat the short-run interest rate as a strongly exogenous policy variable, meaning that there is no reaction function in the model linking the inflation forecast to the interest rate. The important monetary feedback variable is the exchange rate, determining import prices for given foreign prices. The exchange rate depends on inflation, the short-run interest rate and foreign variables.
Section 9.2 sets out the core model of inflation as a wage-price system, conditional on output, productivity, unemployment, and the exchange rate. After evaluating steady-state properties, we derive a dynamic model for wage and price growth. We enlarge this core model to include relationships for output, productivity, unemployment, and exchange rates in Section 9.3, and the exogeneity assumptions underlying such a modelling strategy are examined in Section 9.4. Equipped with the core model and the marginal models we next establish a small econometric model. Despite aggregation of aggregate demand, it is seen that the simultaneous model captures essential features of the transmission mechanisms in the inflationary process for the small open economy. It provides a testing bed for the impact of policy changes on the economy. In particular, it highlights the behaviour of exchange rates, which is central to the conduct of monetary policy in small open economies. The exchange rate behaviour is characterized by a data-consistent empirical model
with short-run interest rate and inflation effects, and convergence towards purchasing power parity (PPP) in the long run.
Section 9.5 contains a discussion of the main monetary policy channels in the model, that is, both the interest rate and the exchange rate channels. We also evaluate the properties of the model for inflation forecasting, while we study the effects of an exogenous change in the interest rate in Section 9.6. In Section 9.7 we sum up our experiences so far.