Read All About It: Bloomberg News and Google Data to Trade Risk
11 Pages Posted: 23 May 2014
Date Written: September 4, 2013
Abstract
In this paper, we construct indices to measure sentiment using Bloomberg news data and also data based on Google searches (Google Domestic Trend indices). We find that in many instances, Google Domestic Trend data displays significant seasonality, which needs to be removed to create a more useful signal. Our Google Shock Sentiment index uses information related to searches on unemployment and bankruptcy, whilst our Bloomberg bull-bear index encompasses the relative number of bullish versus bearish Bloomberg news articles. We use both indices to filter exposure to S&P500 and G10 FX carry. We find that for both assets, the Google based trading rules significantly cut drawdowns and improve risk adjusted returns compared to long only exposure, whilst for our Bloomberg based rule, it seems to add value mostly for G10 FX carry. Our S&P500 Google based trading rule has risk adjusted returns of 0.74 and drawdowns of -18% since 2005, compared to risk adjusted returns of 0.27 and drawdowns of -57% for the long only rule. Our G10 FX carry BBG based trading rule has risk adjusted returns of 0.48 and drawdowns of -17%, compared to risk adjusted returns of 0.21 and drawdowns of -35% for the long only rule. Our Google filtered G10 FX carry rule similarly performs well.
Keywords: foreign exchange, news analytics, Google Trends, Bloomberg News, systematic trading
JEL Classification: F31, E32
Suggested Citation: Suggested Citation