 |
The quality of forecasters’ inflation forecasts is relatively low, and generally biased upwards, nevertheless, the accuracy level of most of forecasters is higher than the inflation expectations derived from the market, measured as the gap between the yield to maturity of nominal and CPI linked bonds. |
| |
A better inflation forecast - the optimal forecast - was found to be the simple average of all the private forecasters. |
| |
Private-sector economists who make accurate forecasts of inflation in the short term are not necessarily the ones who make good forecasts of long term inflation. |
| The Bank of Israel today published a new research paper by Mickey Blank that examines regular inflation forecasts by private-sector economists. Such private forecasters’ inflation forecasts are one of the important indicators examined by the Bank of Israel in its attempts to understand the inflation environment and in its conduct of monetary policy. Hence an analysis and understanding of these forecasts is important to enable an assessment to be made of the quality of the forecasts received by the Bank of Israel. |
| In this study the author examined the forecasts submitted to the Bank by seven forecasters in the period from July 2002 to April 2007. The forecasters were graded within each of four different forecast horizons, from one month to twelve months, according to the indices normally used in the literature that retrospectively measure the deviations of the forecasts from actual inflation. The results show a lack of consistency in the ability of forecasters to predict inflation in the different periods––no forecaster was found who consistently outscored the others throughout the period in all the forecast horizons. Those who successfully predicted inflation in the short term were not necessarily those who did so for the twelve months forward. |
| The average inflation forecasts of all forecasters, all along the sample's period and in all the forecast's ranges, is higher than the inflation actually observed, that is, there is an upwards bias in the inflation forecasts. Furthermore, the accuracy of forecasts measured by the root mean squared error (the difference between values forecast and the values actually observed), is relatively low - 0.26 percent for one-month forecasts and 2.4 percent for twelve-month forecasts. In a comparison between forecasters’ forecasts quality regarding the inflation expectations derived from the market it was found that the accuracy level of most of forecasters is higher than the inflation expectations derived from the market. |
| The study includes a review of the main literature in assessing and grading the quality of forecasts. Many studies claim that by combining the predictions of different forecasters, the information on the future change in the predicted variable can be improved. Those papers deal with the question of what is the optimal combination of forecasters that yields the most accurate forecast of the variable being predicted. |
| The current study found several optimal combinations of forecasts by the different forecasters using different weights for the forecasters, as opposed to the simple average calculated by giving equal weight to each of the forecasts. Although these combinations reduce the deviations between the forecasts and actual inflation, statistical tests carried out on the results show that these combinations are not significantly better than the simple average, and it is therefore recommended that the simple average of all the forecasters be used as the indicator of future inflation. |
| Another finding of the paper relates to the information contributed by the forecasters: the predictions of all the forecasters contain added information and contribute to the average. Therefore no forecaster should be omitted from the average. |