Autor: Llada Martín*, Aromí Daniel**
Institución: *IIEP-BAIRES (UBA-CONICET), **IIEP-BAIRES, FCE-UCA
Año: 2023
JEL: E0, E3
Resumen:
We use mass media news to generate a set of quantitative indexes related to the inflation rate in Argentina for the period 2017-2023. We distinguish the information content of different features of mass media news. First, an indicator of attention allocated to inflation is generated. Second, a set of indexes which captures signals about inflation direction is built. In-sample and out-of-sample exercises confirm that using mass media content leads to gains in forecast accuracy. We find that indexes that capture inflation direction achieve the strongest accuracy gains. The proposed indicators compare favorably with traditional macroeconomic indicators and those based on social media. Combining mass media and social media leads to interesting information gains. Furthermore, the forecast accuracy of indexes based on NLP techniques outperforms traditional approaches.