Machine-Coded News-Based Sentiment Index for Turkey and its Impact on Exchange Rates

Abstract

This paper, initially, develops an automated system that conducts a content analysis using newspaper coverage to generate a high frequency news-driven sentiment index. The system classifies the news as good, bad, or neutral depending on word frequencies. Then, the study investigates the relationship between the news-based sentiment index and exchange rates in Turkey after controlling for both domestic and foreign macroeconomic fundamentals. The results reveal that US economic announcements drive the value of Turkish Lira whereas the Turkish economic announcements and sentiment index are ineffective. When other financial variables are considered, there is strong evidence that the sentiment index affects the stock market returns. Turkish macroeconomic announcements are insignificant in the stock and bond market specifications. The US announcements affect medium to long-term bond yields.