engleski

Assessing the quality of real growth rate of GDP and inflation forecasts for the Croatian economy

The paper analyses the quality of the real growth rate of GDP and inflation forecasts for the Croatian economy. Thus far only the research by Krkoska and Teksoz (2007) has included data regarding Croatian economy but the research was based on EBRD’s GDP growth forecasts and included 25 transition countries from Central and Eastern Europe and former Soviet Republics with the results only being reported at the aggregate level as the countries were grouped in three sub-regions. Recent research by Fioramanti et. al. (2016) which analysed the European Commission’s forecasts did not include Croatia because of lack of data. Therefore, in this research for the first time analysis of forecast data related to the Croatian economy is performed and the results are reported in detail. The analysis is based on forecast data of 6 different institutions belonging to the private, public, domestic and international sectors for the 2006-2015 period. The conducted tests include accuracy, biasedness and efficiency tests. Regarding biasedness and efficiency we pursue the econometric framework first presented by Davies and Lahiri (1995) for the tree-dimensional panel dataset. Accuracy tests include the Pesaran and Timmermann (1992) directional accuracy test. Research results regarding accuracy for both variables generally do not exhibit anomalies and show a rise in forecast error (measured by MAE) with the increase in forecast horizon. Also, using MAE we find that forecast accuracy can be compared to that of EU member states, i.e. accuracy for both variables falls in top 50% of EU member states. Furthermore, as for other economies subject to similar research we find inflation forecasts to be more accurate than the GDP forecasts which can be explained by the fact that it takes time for national statistics offices to come up with accurate GDP data and by clear and stable monetary policy framework. Directional accuracy tests show that forecasters are able to predict the sign of the change in variable (increase vs. decrease) for shorter forecast horizons but find much more difficult to predict the change in intensity (acceleration vs. deceleration). Regarding GDP all the forecasts are biased since all forecasters over-predict GDP growth rate in the analysed period (even if they accurately predict a fall in GDP) and sometimes inefficient implying significant lack of rationality. Inflation forecasts exhibit pronounced bias only in the case of 2 out of 6 institutions. This is accompanied by a solid performance in efficiency suggesting a relatively high degree of rationality in inflation forecasting providing further evidence that inflation forecasting presents much less of a challenge.

GDP growth rate, inflation, forecasting, tree-dimensional panel

nije evidentirano