An interest rate policy is symmetrical if the nominal interest rate response to equally large positive and negative deviations from the inflation target are of identical magnitude but with opposite signs. The governor of Norges Bank (NoB, The Norwegian central bank), Svein Gjedrem, stated in 1999 that good monetary policy includes a symmetrical interest rate: “Over time it is important that the interest rate is set symmetrically. Such symmetry is necessary to sustain the expectations of nominal stability” (Gjedrem, 1999 ). Figure 1.Norges Bank’s expectation of the sight deposit rate in the baseline scenario with fan chart. Per cent. Quarterly figures. 04 Q1 – 08 Q4. Source: Norges Bank, Inflation Report 3/2005.
Norges Bank began its inflation targeting policy about two years after Gjedrem made this statement. The governor has not spoken in terms of a symmetrical interest rate during the period of inflation targeting. Implicitly however, NoB has communicated a symmetrical property of the interest rate by the use of fan charts (see Figure 1). The purpose of the fan chart is threefold: it should provide a forecast of the most likely outcome of the economy. The most likely outcome is represented by the thick black line at the centre of the projections. Second, it should convey the degree of uncertainty surrounding the most likely outcome. The degree of uncertainty is represented by the width of the fans. Lastly, it should provide information on the balance of this uncertainty. The spread around the most likely outcome should indicate the risks related to the level of uncertainty. As illustrated in Figure 1, the risks of an interest rate lower than the most likely outcome is equal to the risk of an interest rate higher than the most likely outcome. The balance of risk is symmetrical.
The optimality of a symmetrical interest rate is based on a simplified model of the economy, often referred to as the linear-quadratic framework. There is however reason to question if the linear-quadratic framework serves as a good model. If any of the assumptions underlying this framework is changed, it will directly impact interest rate setting and a symmetrical interest rate is no longer optimal. The expressed view of a symmetrical interest rate makes thus a tight restriction on monetary policy.
The purpose of this thesis is twofold: it will demonstrate why the property of a symmetrical interest rate / symmetrical fan charts makes a tight restriction on monetary policy. Second, it will test if one of the necessary restrictions for a symmetrical interest rate to be optimal can be rejected on Norwegian data.
The thesis proceeds as follows. The first chapter begins by addressing the questions of what monetary policy is, which variables the central bank wants to stabilise and how it can achieve its goals by using the interest rate as its instrument. With respect to the change towards inflation targeting, the economic rationale when choosing a precise inflation target and a production target, is discussed. This discussion makes the foundation of the economic model derived at the end of the chapter. The model is based on the linear-quadratic framework. The purpose of the first chapter is to illustrate how the linear-quadratic framework leads to an optimal symmetrical interest rate. The second chapter questions the symmetrical interest rate as an optimal result. It thoroughly analyses how altering some of the assumptions made in chapter one impacts the interest rate setting. Specifically, it is the shape of the loss function, the shape of the aggregate supply curve, whether shocks are additive only and the credit channel that are discussed. The third chapter tests if one of the necessary conditions for a symmetrical interest rate to optimal, the linear aggregate supply curve, can be rejected on Norwegian data. The supply curve is taken from one of NoB’s economic models. It will be shown that linearity of the supply curve cannot be rejected. The supply curve has however poor empirical properties, suggesting that the lack of evidence of non-linearity not necessarily is evidence of linearity. Another explanation could be that the model is poorly specified. Estimation is done using Eviews 5.1.