Sep. 26 2018 — Investors sometimes view sentiment signals as interchangeable: one indicator is the same as the next. Our research shows that this is far from the case. In this report we show that: a) sentiment-based signals from different types of market participants (management, analysts, hedge funds) have each been historically predictive; b) such signals have low correlations with each other; and c) combining sentiment indicators in a simple two-factor framework produces historically strong results. In fact, sentiment as a “style” of investing compares favorably to other quantitative investment styles over our test period (see graph in full report).
Our findings for the Russell 3000 include:
- Companies where management is both positive/optimistic and fact-focused outperform historically. Combining two earnings call transcript factors – percent positive words and numbers to words – produces a strategy with an annualized long-only active return of 5.7%, a hit rate of 72%, and an information ratio of 2.0.
- Hedge fund sentiment confirms and complements management sentiment. Combining management percent positive words, from earnings calls, with a hedge fund strategy (ownership level minus short interest) results in an annualized long-only active return of 5.1%, a long-short return of 10.7%, and a long-short hit rate of 75%.
- Market sentiment surrounding earnings calls amplifies the effectiveness of earnings transcript-based signals. A blend of management percent positive words and earnings announcement return results in a 5.3% annualized long-only active return, a 71% hit rate, and a 1.8 information ratio.
- Analyst sentiment, as reflected in target price/recommendation changes, adds an important voice to ownership-based signals. Combining analyst recommendation change with a hedge fund ownership strategy results in a 5.0% long-only active return, an 10.9% long-short return, and a long-short information ratio of 2.2.
Their Sentiments Exactly: Sentiment Signal Diversity Creates Alpha Opportunity
Download The Full Report
Learn more about Market IntelligenceRequest Demo
Natural Language Processing, Part I: Primer
Natural Language Processing – Part II: Stock Selection