Calculating news sentiment indexes at the company level in real time offer a better understanding of the role of sentiment in asset pricing.
Below are some of our key results.
Company Sentiment Holds Strong Predictive Power
- The return spread between positive and negative sentiment changes is significant over a 2 day horizon.
- The return spread is consistently positive over time and across industries covering the test period of 2000 – 2011.
- More extreme sentiment changes result in greater positive or negative price impact.
Company Sentiment Strategies Deliver High Risk-adjusted Performance
- On average, the return spread between companies with positive and negative sentiment changes is about 1.05% intraday, while the effect over days 1 and 2 is 0.34% - improving to 0.63% when only considering the most extreme sentiment stocks.
- Based on a universe including over 4,000 US companies, going long stocks with positive sentiment and short stocks with negative sentiment generates Information Ratios of 2.0 - 3.0 with daily Hit Ratios of almost 60%.
- On average, the strategy generates a total of 150 daily trading signals (long and short) resulting in average portfolios of around 300 stocks.
In recent years, quantitative investors have been struggling to come up with new sources of alpha. The increased “quantcentration” in the investment space has made most of the traditional quant factors less effective, and hence the “hunt” for new alpha has become even fiercer. However, the development of behavioral finance theories and historical anecdotes like the dotcom bubble suggest that investor sentiment may be an untapped source of alpha and has demanded increased attention from the investment community. Theoretically sentiment is usually introduced into asset pricing through the effect on risk or return expectations, which cause asset prices to diverge from their intrinsic value for certain periods of time until fully offset by rational investors.
Although the role of sentiment is drawing more and more attention from practitioners and academics, most of the research is done at the market level due to limited availability of security level sentiment indicators. However, with the development of RavenPack News Analytics, a real–time news analytics service that covers more than 28,000 publicly traded companies worldwide; it’s now possible to quantitatively construct a news sentiment indicator at the company level (in real time) and thereby obtain a better understanding of the role of sentiment in the pricing of individual assets. To incorporate news analytics into a stock selection process, one of the challenges is to identify a methodology that works well for constructing a company sentiment indicator.
The objective of this study is to address this issue and define a methodology that captures potential sentiment shifts important to asset pricing. To this end, we construct several candidate indexes based on various news sentiment analytics, focused on event or story-level sentiment, different calculation methods, and different aggregation windows. In the next section, we will briefly describe the construction of the RavenPack Company Sentiment Index. In Section 3, we discuss the statistical significance of the various index candidates when it comes to explaining daily stock returns in a regression framework.
Section 4 presents the future return pattern following positive and negative sentiment changes. Furthermore, we consider the return spread consistency across time and industries, as well as address the market cap characteristics across sentiment deciles for the two calculation methods. Section 5 illustrates the economic significance of a trading strategy based on sentiment index changes and Section 6 reveals our conclusions.
Click here to download the PDF and continue reading the "Short-Term Stock Selection Using News Based Indicators" White PaperRequest White Paper