The NBR uses a structured forecasting framework for its quarterly macroeconomic projections. This framework refers to

1. database management and
2. operative and continuous development of appropriate models for analyses and forecasting.

Models operated at NBR can be classified into three distinct groups. All of them are interconnected and thus the final outcome - the quarterly macroeconomic forecast (or projection) - integrates the information provided by all models.

• The first group is made up of models based on empirical econometric estimations between dependent macroeconomic variables (e.g. inflation, the growth rates of GDP and GDP components, lending and deposit rates, labour market indicators) and their economic determinants - independent variables (e.g. exchange rate, oil price, interest rates or real wage).
• The second group refers to the central model (the model for medium term analysis and projection - ro. MAPM) designed to explain the dynamics of relevant macroeconomic variables at business cycle frequencies. The model is made up of behavioral equations that describe relationships between key macroeconomic variables (e.g. inflation rate, interest rate, exchange rate, output gap). Their dynamics is derived according to an equilibrium based on optimal (rational) behavior of representative economic agents (such as households, firms, financial institutions, the central bank, the external sector).
• The third group includes satellite models (for GDP components, labour market and balance of payments) which can supply additional information as well as provide a check of the consistency of the final projection.

The process for generating quarterly macroeconomic projections consists of three steps, based on a rigorous timetable which is revised each year. These steps are, in chronological order, the following:

• Near term forecasts - mainly the outcome of empirical models (the first group of models listed above), which have a very good forecasting performance up to a several months' horizon. These produce forecasts of CPI and GDP components, key interest rates for the transmission mechanism as well as some labour market indicators. Usually these forecasts are relevant for one or two quarters ahead. Expert projections are produced for variables that cannot be modelled and estimated using econometric methods (e.g. changes in administered energy prices or adjustments of excise duties for some products).
• Estimating the position of the economy over the business cycle (initial starting conditions). This is the first step in integrating the above near term forecasts and the expert projections within the common framework of the central model. For this purpose, it is necessary to decompose real economic variables into a trend and a cyclical component (gap or deviation from trend). This is done using a statistical multivariate filter (Kalman filter). The filter's consistency with MAPM is assured by incorporating the same transmission mechanism into both components. The analysis is supported by additional information provided by a comprehensive set of leading and coincident indicators of the business cycle.
• The main forecast using the model for medium term analysis and projection (MAPM). The model incorporates a stylized version of the main monetary policy transmission mechanisms (Monetary policy transmission mechanism). The model is built, developed and calibrated with the purpose of capturing the dynamics of macroeconomic variables over the business cycle. At the same time, over the long term, the model converges to a well defined steady-state. The projections presented in the Inflation Report actually comprise only 8 quarters, which is in line with the empirically estimated relevant horizon required for monetary policy to wear off most of its impact on the economy. During this time horizon, the model's specific mechanisms are supposed to work by adjusting macroeconomic variables towards their medium-term equilibrium.

The model for medium term analysis and projection (MAPM) integrates all the relevant information and at the same time provides the consistency needed in analyzing, communicating and implementing monetary policy decisions. Nevertheless, a projection is not the outcome of an automatic procedure but an iterative process, which takes into account and incorporates the views of both the staff and the bank Board. The forecasts generated by the model are a very important input in the NBR Board's policy decision making.

The model for medium term analysis and projection (MAPM) integrates all the relevant information and at the same time provides the consistency needed in analyzing, communicating and implementing monetary policy decisions. Nevertheless, a projection is not the outcome of an automatic procedure but an iterative process, which takes into account and incorporates the views of both the staff and the bank Board. The forecasts generated by the model are a very important input in the NBR Board's policy decision making.

Macroeconomic modelling activity within the National Bank of Romania primarily aims at supporting monetary policy decisions by ensuring the necessary theoretical and empirical tools. The bulk of scientific research activity is focused on improving the quarterly model for medium term analysis and projection (MAPM), further developing sectorial satellite-models, as well as making various analyses supportive of the model’s main outcomes. For example, a DSGE (Dynamic Stochastic General Equilibrium) model with partial euroization was developed and estimated for Romania. The DSGE model has a superior analytical structure as compared to MAPM and thus it is used via shadow forecast exercises to supply additional information within the monetary policy decision-making process. Details on the research activity are presented in “Economic research” chapter of the annual report.