The paper proposes new text-based indicators of sentiment and economic policy uncertainty (EPU) for Italy with a daily frequency. These indicators are built using textual data from around 1.6 million articles published in national newspapers. The authors adopt an innovative approach based on a dictionary that takes into account the context in which the words are used and their role as adverbs or negations which change the meaning of the sentence (from positive to negative or vice versa).
To evaluate the utility of the proposed text-based indicators of sentiment and economic policy uncertainty (EPU) both for the Italian economy as a whole and for specific sectors or themes, we performed a number of forecasting exercises to predict the short-term behaviour of some macroeconomic variables. Our results show that employing text-based indicators of sentiment and EPU reduces the uncertainty of monthly predictions for Italian economic activity, especially during recessions, and improves the accuracy of the weekly estimate of Italy's GDP.
Forthcoming in: International Journal of Forecasting