Starting from 2008, instead of providing a verbal description of the expected interest rate path the CNB will publish its interest rate forecast in numerical form, as a fan chart. This box explains the conditional nature of the rate forecast and the methodology for the construction of fan charts.
By releasing its interest rate forecast the CNB is continuing to enhance its monetary policy transparency. If the public can better comprehend the central bank’s actions and assess the quality of its analyses and forecasts, its trust in the bank’s ability to keep inflation on target increases. It is vital, however, that all users of the central bank’s forecasts are aware that the published forecast-consistent interest rate path should in no way be interpreted as a commitment of the central bank to set interest rates in line with the forecast. There are two reasons for this. First, the forecast represents the most probable future path of interest rates under given initial assumptions and information. New information on the domestic and global economy that comes in after the forecast is drawn up can change the interest rate outlook. The second reason is that the CNB Bank Board may not entirely agree with the forecast prepared by the Monetary and Statistics Department or may regard the associated risks as being skewed to one side or the other. Chart 1 shows that interest rates have often deviated from the forecasts in the past as well. The chart illustrates all the projected interest rate paths from individual consecutive forecasts as well as the actual interest rate path since 2002. The interest rate shown is the three-month money market rate (3M PRIBOR), which is used in the CNB’s forecasting system to proxy for the CNB’s key monetary policy rate, i.e. the two-week repo rate.
To emphasise uncertainty, or the conditional nature of the forecast, the CNB has started using fan charts. Starting with this Inflation Report, fan charts are used to illustrate the uncertainty surrounding not only the interest rate forecasts, but also the outlook for headline inflation, monetary-policy relevant inflation and GDP growth. The central line in each chart represents the baseline scenario of the forecast, and the bands around this line illustrate the uncertainty surrounding the forecast. The successively lighter bands represent widening confidence limits. The darkest band around the centre shows the development which can occur with 30% probability, while the other bands successively show the developments occurring with 50%, 70% and 90% probability.
The confidence intervals are constructed with reference to the errors of previous CNB forecasts for the relevant variables. The advantage of this approach is its simplicity and clarity and the fact that it captures all the types of uncertainty contributing to the non-fulfilment of the forecasts (the uncertainty regarding input assumptions as well as the settings of forecasting instruments). A potential disadvantage is that in the event of changing conditions and economic shock strengths the past prediction errors may not precisely capture the size of the uncertainty going forward. The confidence intervals should thus be regarded as only tentative.
When creating the confidence intervals in fan charts, we assume a symmetric normal distribution around the baseline scenario of the forecast. The variance of this normal distribution is based on the Root Mean Square Error (RMSE). The RMSE for a given forecast horizon is defined as follows:
where X is the actual value of the forecast variable, F is the forecast for this variable drawn up in the past with a corresponding forecast horizon, and the difference (X – F) is the forecast error. The RMSE is calculated from the errors of the last eight available forecasts, fulfilment of which can be assessed at the eight-quarter horizon. The RMSE for the individual forecast horizons is then further smoothed linearly so that the probability fan has a smooth shape. The RMSE is updated once a year. The confidence intervals are derived for the individual horizons and confidence levels from the relevant (smoothed) RMSE and the relevant quantiles of the normal distribution.
The confidence interval for the interest rate and GDP growth paths widens linearly over the entire forecast horizon, whereas that for headline and monetary-policy relevant inflation widens only for the first four quarters and then remains constant. This is consistent with the evolution of the CNB’s forecast errors to date and generally also with the inflation targeting regime, under which the central bank anchors inflation close to the target at the medium-term horizon, whereas the interest rate and GDP paths reflect the effects of unforeseen shocks faced by monetary policy.