Publikationen des DIW Berlin

Abstract:

In this paper, we examine the predictive ability of automatic and expert-rated media sentiment indicators for German inflation. We find that sentiment indicators are competitive in providing inflation forecasts against a large set of common macroeconomic and financial predictors. Sophisticated linguistic sentiment algorithms and business cycle news rated by experts perform best and are superior to simple word-count indicators and autoregressive forecasts.

JEL-Classification:

C53;E31;E37

Keywords:

Inflation prediction, media sentiment indicators, news reports, real-time forecasting