Fiscal Forecasting in Italy
- Autores: Laura Carabotta
- Directores de la Tesis: Elisenda Paluzie i Hernández (dir. tes.), Peter Gunther Claeys (dir. tes.)
- Lectura: En la Universitat de Barcelona ( España ) en 2015
- Idioma: inglés
- Tribunal Calificador de la Tesis: Raúl Ramos Lobo (presid.), Javier José Peréz García (secret.), Roberto Patuelli (voc.)
- Materias:
- Enlaces
- Tesis en acceso abierto en: TDX Dipòsit Digital de la UB
- Resumen
The thesis “Fiscal forecasting in Italy” is comprised of three main chapters in which is analyzed, from an empirical point of view, several issues related to public finance forecasts, with an application to Italy. Chapter II, “Accuracy of fiscal forecasts in Italy” is focused on one of the most important aspects of the new Treaty: it requires that the decisions and recommendations taken by the European Commission are no longer be based on outcomes but on forecasts. In this chapter, I evaluate whether fiscal forecasts for Italy are accurate and econometrically efficient. I focus on a large number of deficit forecasts for Italy that come from a variety of sources, including both public and private agencies as well as Italian and international institutions. I analyse the extent of the discrepancies between the yearly released deficit on GDP and its forecast in Italy from 1/1992 to 12/2011. I conduct two types of analysis. In the quantitative analysis, I carry out different accuracy tests to detect which organization is the best forecaster and in what part of the year better results are published. I also compare forecasters’ performance against a naïve benchmark model, which provides a minimum level of accuracy. In the qualitative analysis, I consider the quality of the forecasts and I test efficiency, unbiasedness and serial correlations. I conclude that all fiscal forecasters for Italy provide unbiased and inefficient forecasts. In general, forecast errors do not persist in a regular way. The most relevant result of this analysis is that private forecasters are frequently more accurate than national and international ones. In Chapter III, “Combine to compete: improving fiscal forecast accuracy over time”, take advantage of the information contained in all individual budget forecasts analysed in the previous chapter to improve their accuracy. I do this by projecting combined forecasts through pooling the judgment and expertise of the forecasters. Following this idea of improving the forecasting accuracy, I apply a variety of combination techniques, both simple and advanced, which account for past forecasting performance, to compute a combined forecast. I look at a same dataset which is analysed in the previous Chapter. My main finding is that different combinations of budget forecasts often result in more accurate forecasts than individual models. This is particularly the case for a weighted forecast combination and Rbest that value the forecasts that have been more accurate in recent periods. Standard tests of forecasting accuracy show that even one year ahead, some of the pooled forecasts significantly outperform a naïve model. I use recently developed fluctuation tests to check forecasting accuracy over time I find that the weighted forecast combinations outperforms other predictors overall years. Its improvement in accuracy is statistically significant when compared to a naïve model. Chapter IV, “Nowcasting public finance in Italy,” moves from the idea of forecast and combination of annual data to the most recent idea of nowcasting fiscal variables. The reason is to give policy makers the capacity for dynamic monitoring of the public budget’s cash flow. This monthly analysis exploits the information at higher frequencies before the official figure becomes available. The approach that I use consists of using different nowcasting techniques that are well known in the literature. In particular, I propose a set of models that are parsimonious and suitable for real-time monitoring of the fiscal deficit. I conclude that the linear regression models outperform the other techniques used. The introduction of public finance and economic confidence variables and Google trends results in performance gains when compared with the VAR, the time series and autoregressive models.
