Trading Relative Value Based on News Indicators

RavenPack | August 02, 2013

In this study, we develop a market level sentiment indicator showing that companies subject to negative sentiment contain more information on the general market environment than companies subject to positive sentiment.

Executive Summary:
We evaluate our indicator by constructing a simple investment strategy that goes long or short the return spread between the S&P500 Index versus the MSCI ex. U.S. index when their sentiment spread is positive or negative, respectively.

Our study demonstrates that sentiment spread shows strong predictive power over longer investment horizons. The strategy not only seems interesting from a relative value point-of-view, but also from a tactical asset allocation perspective.

Companies with negative sentiment contains more information on market environment than companies with positive sentiment

A relative value strategy with a 1-month holding period provides a 50% improvement in Information Ratio when focusing on companies with negative rather than positive sentiment.

Sentiment spread shows strong predictive power over longer investment horizons

A strategy that goes long or short the return spread when the sentiment spread is either positive or negative, respectively, has yielded an Information Ratio of 1.0 with an average holding period of 10 months.

1. Introduction

In this paper we suggest an alternate approach to building market-level sentiment indicators than our previous method of aggregating the counts of company related positive and negative news events over a given time horizon. Here, we construct a set of indicators that focus on different subsets of our company universe based on the percentile distribution of RavenPack's Company Sentiment Indicator (CSI). So, instead of considering sentiment across all companies for a given country or region, we hypothesize that companies of a particular CSI percentile are more responsive to changes in the market environment than others - caused by an expansion and contraction of information over time.

When considering news about companies, we typically find a strong positive sentiment bias. It is only when sentiment is at its worst that the ratio of positive to negative news gets close to 1. Such changes seem to be driven not only by an increase in the amount of negative news, but also by a drop in the amount of positive news flowing to the market. Due to information expansion and contraction, we find it reasonable to believe that more information about the market environment can be found amongst certain subsets of our equity universe. Given that such effects are most pronounced during market turmoil, on average, most value is likely to be found by tracking companies with negative sentiment rather than positive - something we'll address as part of this paper.

Figure 1 below plots the cumulative change in the count of positive and negative news across U.S. companies. We highlight two periods where the difference in the amount of negative to positive news goes through a period of expansion and contraction. One period relates to the Dotcom bubble and the second to the recent Financial Crisis