Sentiment Reversals as Buy Signals

Generally, the use of news analytics falls into three main categories including (1) triggering events for high and low frequency strategies, (2) quant factors, and (3) Sentiment Indices/overlays. Previously, I have presented studies conducted by Macquarie Equity Research showing how news derived quant factors can add value to existing multifactor models. Also, I have shown how one can construct a News Sentiment Index using a bottom-up approach. In this posting I will present an example of how a sentiment reversal can be considered as buy signals (a triggering event).

In a recent study by John Kittrell from Knightsbridge Asset Management, a sentiment reversal is defined as a news story that causes a company to emerge from a prolonged period of negativity, as determined by at least 30 negative net sentiment measurements over at least 30 calendar days, see Figure 1.


Monthly measurement periods are used in order to increase the likelihood that the reversals are non-accidental. To address strategy scalability, sentiment reversals apply only to companies with a minimum market capitalization of $500 million (market growth adjusted). Finally, the out-of-sample period covers the period 2000 through 2008.

On average, the reversal events generate a 16% annualized excess return based on a holding period of one year, but in order to form a more a realistic setting, a 10-stock Monte Carlo study is considered, with starting points taking place pre-2000 (Figure 5 below).

In one scenario, it is only possible to enter a new position when cash is available and only when the reversal event is identified. While in the second scenario, it is possible to enter a position up to four months after the reversal event. Hence, less periods with pure cash positions are observed under this scenario. On average, waiting up to four months to enter the market after an event results in a lost opportunity of about 2%.

Some of the key findings from Knightsbridge’s study include:

• More sentiment reversals are present in bull markets than in bear markets
• Sentiment reversals tend to do better in bull markets, while still outperforming the market in bear markets
• On average, the reversal events generate a 16% annualized excess return based on a holding period of one year
• The performance of the different 10-stock Monte Carlo strategies, very much depends on the starting point and is reflected in excess returns to the S&P500 ranging between 4-11% with an average of about 7% over the test period
• All portfolios outperform the S&P500 over the test period

For this study, Knightsbridge Asset Management used news sentiment data from Dow Jones and RavenPack. Download the full paper here.