The role of assessments and judgement in the use of the macroeconometric model RIMINI.

• Author: Olsen, Kjetil
• Position: Model developed by Norway's Norges Bank Research Department

The Inflation Report's projections for economic developments are important for Norges Bank's conduct of monetary policy. The macroeconomic model RIMINI is used as a tool in developing these projections. This article provides insight into key aspects of the model's description of the inflation mechanism and how the model is used to make economic forecasts. Assessments and judgement play an important role in this work. The assessments are partly based on systematic analysis of current statistics and information from other models that shed light on temporary factors. Historical forecast errors also provide insight into the use of the model. Knowledge of this kind provides the basis for steering the model in the forecast period. The article also illustrates how the model may be used to study effects of interest rate changes.

1 Introduction

Norges Bank's projections are based on analyses of the most significant relationships in the economy and on key assumptions about economic policy and international conditions. The projections reflect an overall assessment of economic developments. Norges Bank also analyses the effects of monetary policy and the impact of various shocks on the Norwegian economy. Such shift analyses are published regularly in the Inflation Report.

The macroeconomic model RIMINI is an important forecasting tool in this work. Smaller models that have been developed to study special issues are also used. The results from this type of analysis are incorporated in the RIMINI model. Thus, the projections published in the Inflation Report express an overall assessment of the results from different models and current developments in the Norwegian economy. This article looks more closely at our use of the RIMINI model for projections and analysis, with special emphasis on price and cost inflation and the effects of monetary policy.

RIMINI is a macroeconomic model that has been developed by the Research Department of Norges Bank. The model takes account of many of the most important relationships in the Norwegian economy and explains both real and nominal variables. It combines and takes account of empirical and theoretical knowledge about these relationships as they have functioned in the past and contributes to a consistent analysis of the interaction between them. Using a set of assumptions about future economic policy, among other things, the model provides quarterly projections for developments in the Norwegian economy.

The RIMINI model does not necessarily reflect Norges Bank's view of the economy. However, the model and our use of it provide the basis for the projections and especially for assessing how changes in the assumptions may affect the projections. The model attempts to explain the main systematic features, but not every detail of economic developments. Therefore, as we work on the Inflation Report, the model is changed regularly. In addition, we frequently use information from other models or from current economic developments. Thus, the projections in the Inflation Report are not merely a result of the RIMINI model's properties. The assessments of model users are equally important.(2)

Section 2 presents a brief overview of the RIMINI model's scope, structure and background data. Section 3 looks more closely at price and wage formation in the model and how the model is used. Section 4 discusses how we use the model to study the effects of interest rate changes. Section 5 provides a summary.

2 General information about the RIMINI model

The RIMINI model is designed to make short and medium-term projections for the Norwegian economy as well as for policy analyses. Within a one-year time horizon, current developments in the Norwegian economy as they are presented in monthly statistics play a prominent role in preparing the projections. For medium-term projections, ie from 2 to 5 years, the model's results are used to a larger extent as guidance in making the projections.

Developments in economic variables depend on a number of mutually dependent mechanisms that are often complex and difficult to quantify. Overall demand, for example, affects both activity level and employment, which in turn determines income levels, which again affect overall demand. Changes in real variables affect nominal prices for goods, services and labour. Therefore, in the RIMINI model, the endogenous variables are determined in a simultaneous system of equations.

The interest rate functions as a monetary policy instrument and is therefore a key exogenous variable in the model. A technical assumption underlying the Inflation Report's baseline scenario is that interest rates follow expectations in the money and bond markets as reflected in forward rates. Projections are also made on the basis of unchanged interest rates. The exchange rate is also an exogenous variable in the RIMINI model and the baseline scenario assumes that exchange rates remain unchanged. When the model is used to calculate the effect of changes in different variables such as the interest rate, the exchange rate is endogenised by, for example, assuming uncovered interest parity. This will be discussed in further detail later in the article. The RIMINI model contains a good 100 exogenous variables that are not determined by the model and which must therefore be estimated outside the model when making projections. In addition to the monetary policy stance, the exogenous variables primarily describe developments among our trading partners, main word market prices and policy variables that describe fiscal policy.

Expectations about future inflation, demand and other economic variables may affect household and corporate behaviour. Expectations formation is not explicitly modelled in RIMINI, but the model contains a number of explanatory variables that may capture economic agents' expectations. To take account of adjustment lags in the economy, the model has been given a dynamic specification where lagged variables play an important role. Forward-dated variables are not included. This does not imply, however, that the model is inconsistent with forward-looking behaviour.

In the RIMINI model, a main distinction is made between production sectors including manufacturing and construction on the one hand and private services on the other. The former may be characterised as internationally exposed sectors, while the latter is largely sheltered from international competition. The public sector, primary industries and the oil and shipping industries are also represented in the model. Economic developments in these sectors are treated exogenously, in contrast to private services and manufacturing and construction. The model not only reflects conditions in the real economy but also financial and monetary conditions. It also includes an income account for different sectors.

The RIMINI model is based on quarterly data. The quarterly national accounts are the most important data source together with other statistics from the national accounting system and from Norges Bank's database for financial sector balance sheets (FINDATR). Other statistical sources also provide important data for the model. The most recent version of the model (RIMINI 3.14) has been calculated on the basis of national accounts figures in accordance with the European National Accounting System (ENS95).

Like other econometric models, the RIMINI model is changing constantly. New knowledge about methods or economic theory will improve the properties of a model based on research. New observations, evaluation of projections and experience in using the model also provide new insight. Computer tools are under constant development as well, making calculations and simulations more precise and effective. The RIMINI model currently comprises 375 equations, 74 of which are estimated behavioural relationships. These equations will contain add factors that capture the unexplained variation in the left-hand-side variables. Later in the article, we will explain how these add factors may be used when simulating the model to make projections.

The mechanisms and relationships between the model's variables may be regarded as a representation of a large simultaneous probability distribution. However, the number of relationships are too numerous and complex to model simultaneously.(3) Instead, we primarily model single equations separately from the rest. Modelling consists of developing clearly specified single equations where residuals do not contain systematic information that can give the equation increased explanatory power. Further, emphasis is placed on accurate estimation of the parameters and identification of parameters that are likely to be constant over the model's horizon for projections or policy assessments. Finally, the single equations are combined into a complete system.

It is important to use several criteria in evaluating the system's (model's) properties in addition to the properties of the individual equations. First, the individual equations and the model as a whole must be capable of explaining the key features of the data, eg systematic developments and trends in the medium term. The objective is to explain systematic changes in the data and not random variations. Second, the model's long-term equilibrium relationships should be supported by generally accepted economic theory. As the economy is constantly exposed to disturbances, it will seldom be in equilibrium, but there will...